The STAT3–IL-10–IL-6 Pathway Is a Novel Regulator of Macrophage Efferocytosis and Phenotypic Conversion in Sterile Liver Injury

This article is featured in In This Issue, p.883

Phagocytosis is an evolutionarily conserved, multistep process that spans the recognition and engulfment of phagocytic cargo, through cargo processing to Ag presentation (1). Prompt removal of apoptotic and necrotic cells (efferocytosis) is critical to immune tolerance induction and maintenance or re-establishment of tissue homeostasis (2). A failure of efficient efferocytosis has been implicated in the pathogenesis of autoimmune and inflammatory disorders (3–5) such as systemic lupus erythematosus. Chronic inflammatory conditions are characterized by an aberrant release of proinflammatory intracellular components from secondary apoptotic and necrotic cells that fail to be cleared (6, 7). In solid tumors, recognition of apoptotic cells can promote an immunogenic response and antitumoral acquired immune responses (8). Efferocytosis has been postulated to be a mechanism through which initial organ damage, i.e., cell death, programs tissue remodeling, and regeneration. However, the molecular mechanisms underpinning this are not well understood (1, 9); understanding such mechanisms in this process could allow therapeutic targeting. Macrophages are professional phagocytes (10), and they change phenotype depending on microenvironmental cues, which include efferocytosis (11–13). We and others have previously characterized a hepatic restorative macrophage, derived from recruited inflammatory macrophages, which is necessary for tissue remodeling and regeneration following sterile liver injury (14, 15). Furthermore, in liver injury, phagocytosis of necrotic hepatocytes by macrophages prompts Wnt ligand secretion, which promotes liver regeneration (16).

In this study, we have derived macrophages from the bone marrow of control (glycoprotein nonmetastatic melanoma B, Gpnmb⁺) mice or of mice that have a defect in Gpnmb (Gpnmb⁻), an interactor of L chain 3 (LC3). Gpnmb⁻ macrophages will engulf but fail to process their apoptotic cell cargo (17, 18). Macrophages have been fed with apoTs to study the activation of intracellular pathways downstream from the engulfment and digestion of phagocytic cargo. We show that macrophage phosphoSTAT3 (pSTAT3) is a prophagocytic mediator that enhances macrophage phagocytosis, via rapid nontranscriptional regulation of IL-10. pSTAT3 activation is sustained at the later stages of phagocytosis, promoting IL-6 transcription and thereby making efferocytosis more efficient. Following acute and chronic sterile liver injury, Gpnmb⁻ mice exhibited extensive liver damage and increased numbers of proinflammatory macrophages. Treatment with recombinant IL-6 is sufficient to increase efferocytosis in vivo and to promote macrophage phenotypic conversion to prorestorative macrophages.

C57BL/6 mice (CD45.2⁺) were purchased from Charles River U.K. C57BL/6 IL-6 knockout (KO) mice (19) were provided by Prof. S. Anderton, University of Edinburgh. Gpnmb⁺ (DBA/2J-Gpnmb⁺/SjJ) and Gpnmb⁻ (DBA2J) mice were originally imported from the Jackson Laboratory (17). The colony was propagated at the University of Edinburgh. Mice were housed in groups of five or six in open-top cages, and synchronized to a 10–14 h dark/light cycle with access to food and water ad libitum. Mice were bred under specific pathogen-free conditions at the University of Edinburgh. All experiments had local ethical approval and were conducted under U.K. Home Office legislation. Genotyping was carried out by using PCR by TransnetYX.

Wild-type (WT) C57BL6 male mice were allowed to acclimatize for a minimum of 1 wk in a clean animal facility. Prior to carbon tetrachloride (CCl₄) treatment, mice were randomly assigned to treatment groups. Adult male mice (10–12 wk old) were used. Hepatic fibrosis was induced by two injections per week of CCl₄ (0.4 μl/g; Sigma-Aldrich) i.p., diluted 1:3 in olive oil (Sigma-Aldrich) for 6 wk. Animals were culled at stated time points after the final CCl₄ injection.

WT C57BL6 male mice (10 wk old) were allowed to acclimatize for a minimum of 1 wk in a clean animal facility. Prior to acetaminophen (APAP) administration, mice were fasted for at least 12 h. Mice received a single injection (i.p.) of APAP (300 mg/kg) dissolved in warm saline between 23:00 and midnight. Mice were left to recover until morning on a heated mat or in a warm rack; at indicated time points, they were humanely culled according to local ethical guidelines.

Adult male mice at least 8 wk old were used. Acute liver injury was induced by i.p. injection of CCl₄ (0.4 μl/g; Sigma-Aldrich) diluted 1:3 in olive oil (Sigma-Aldrich). Animals were culled at stated time points after CCl₄ injection.

Mice dosed with APAP or receiving a single CCl₄ injection were injected with 100 μl 0.1 mM PKH26PCL diluted 1:10 according to the manufacturer’s instructions (Sigma-Aldrich, U.K.) i.v. via the tail vein to label phagocytic cells. Control mice were injected with 100 μl of diluent only. The percentage of phagocytic cells in hepatic infiltrating and resident is reported. Mice were culled 6 h post-CCl₄ and livers were harvested and processed as described for flow cytometry analysis. Numbers were calculated based on the cell counting performed after isolation of the nonparenchymal cell (NPC) fraction. The percentage of phagocytic cells was calculated on the gate of viable CD45⁺Ly6G⁻CD3⁻CD19⁻NK1.1⁻ cells.

We transferred Gpnmb⁺ CFSE-labeled bone marrow–derived macrophages (BMDMs) and Gpnmb⁻ CMTMR-labeled BMDMs into C57BL/6 mice receiving a single dose of CCl₄. Mice received either Gpnmb⁺ or Gpnmb⁻ BMDMs or both populations in a 1:1 ratio. To avoid an effect of the labeling on the parameter analyzed, the experiment was repeated using Gpnmb⁺ CMTMR-labeled BMDMs and Gpnmb⁻ CFSE-labeled BMDMs. CFSE-labeled or CMTMR-labeled BMDMs were resuspended in Dulbecco’s PBS and administered (2.5 × 10⁶ cells, 100 μl) via the tail vein to mice under gaseous isoflurane/oxygen anesthesia at 2 h after the single CCl₄ injection. Mice were humanely culled by asphyxiation in a rising CO₂ atmosphere. Death was confirmed by neck dislocation. CFSE-labeled or CMTMR-labeled BMDMs were injected 2 h after CCl₄ dosing. CFSE- and CMTMR-labeling efficiency was routinely >90%. Viability posttransfer was routinely >90%. Mice were culled 6 h post-CCl₄ and livers were harvested and processed as described for flow cytometry analysis.

Transferred BMDMs were identified as CFSE⁺ or CMTMR⁺. The percentage of CFSE⁺/CMTMR⁺ cells was calculated on the gate of total viable CD45⁺Ly6G⁻CD3⁻CD19⁻NK1.1⁻ cells. The negative was set on a liver from a nontransplanted mouse. The percentage of Ly6C⁺ and Ly6C⁻ was calculated on the gate of CFSE⁺/CMTMR⁺ cells. The negative was set using the liver from a nontransplanted animal (for single CFSE⁺ and CMTMR⁺) or transplanted with BMDMs labeled with the other tracker (CFSE for CMTMR and vice versa).

Whole blood was collected via cardiac puncture or from the inferior vena cava using 30 μl heparin (200 U/ml) per sample. Blood was centrifuged at 10,000 rpm for 10 s at room temperature in a bench centrifuge (Eppendorf), and plasma was isolated and snap-frozen using dry ice. Plasma was then stored at −80°C and used for protein dosages. Liver tissue was harvested and the left lateral lobe was separated into two pieces and placed in either a freezing isopentane bath or fixed in methacarn for 24 h. The remaining liver and other organs were fixed in formalin (4% paraformaldehyde) for 24 h before paraffin embedding. The central lobe was used for isolation of hepatic NPCs and flow cytometry analysis.

Isolation of the hepatic NPC fraction was performed as described previously (16, 20) with minor modifications. Briefly, mouse livers in situ were perfused with 5 ml 0.9% NaCl solution through the inferior vena cava followed by cutting of the portal vein to remove circulating cells. Livers were then harvested and weighed; the right lobe was homogenized using a scalpel and digested in RPMI 1640 containing collagenase V (0.8 mg/ml; Sigma-Aldrich), collagenase D (0.625 mg/ml; Roche), dispase (1 mg/ml; Life Technologies), collagenase D (1.6 mg/ml; Roche), and DNase I (100 μg/ml; Roche) for 25 min at 37°C, shaking vigorously every 5 min. Digested livers were passed through 70 μm cell strainers, and enzymes were inactivated by the addition of RPMI 1640 with 10% FCS. The NPC fraction containing hepatic macrophages was harvested by two centrifugations at 300 × g, 4°C, 5 s, followed by red cell lysis with 3 ml 1× lysis buffer (BD Pharm Lyse; BD Biosciences) for 5 min on ice. Cells were then counted, and used for flow cytometry.

Nonspecific Ab binding was blocked by incubating cells with 10% mouse serum for 20 s at 4°C, followed by incubation with combinations of primary Abs (each used at 1:200 dilution) for 20 s at 4°C. The following conjugated Abs were used: CD11b BV650 (clone M1/70; eBioscience), Ly-6C V450 (clone HK1.4; eBioscience), CD45.2 AF700 (clone 104; eBioscience), F4/80 APC (dilution 1:100; clone BM8; Invitrogen), Ly-6G PE-Cy7 (clone 1A8; BioLegend), CD3 PE-Cy7 (clone 17A2; BioLegend), NK1.1 PE-Cy7 (clone PK136; BioLegend), CD19 PE-Cy7 (clone 6D5; BioLegend). Cell viability was assessed with Fixable Viability Dye eFluor780 (1:1000; eBioscience) according to the manufacturer’s protocols. After Ab staining, samples were either analyzed immediately or fixed with 10% buffered formalin. Data were analyzed using FlowJo10 software (Tree Star). Hepatic RESIDENT macrophages were defined as viable CD45⁺Ly-6G⁻CD3⁻NK1.1⁻CD19⁻CD11b^lowF4/80^high. Hepatic infiltrating macrophages were defined as: viable, CD45⁺Ly6G⁻CD3⁻NK1.1⁻CD19⁻CD11b^highF4/80^low cells from NPC fraction of digested livers and used to identify macrophage subsets. Subsets were expressed as proportions of total hepatic macrophages or CD45⁺ cells. Quantification of absolute numbers of cells per liver was performed by expressing each subset as a proportion of NPCs, counting total number of NPCs in the digested portion of liver, calculating the total number of NPCs in the whole liver by weight differential, and thus, calculating the total number of each subset. Circulating monocytes (from whole blood diluted 1:1 in heparin) were stained using BD Pharm Lyse (BD Biosciences) before analysis. Monocytes were identified as CD45⁺CD11B⁺Ly-6G⁻CD3⁻Ly-6C^high and Ly-6C^low cells from whole blood and expressed as a percentage of total peripheral mononuclear cells.

Plasma chemistry was performed by measurement of alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), total bilirubin, and plasma (p) albumin. ALT was measured using the method described previously (21), utilizing a commercial kit (Alpha Laboratories). AST and ALP were determined by a commercial kit (Randox Laboratories). Total bilirubin was determined by the acid diazo method described by Pearlman and Lee (22) using a commercial kit (Alpha Laboratories). Mouse plasma albumin measurements were determined using a commercial serum albumin kit (Alpha Laboratories). All kits were adapted for use on a Cobas Fara centrifugal analyzer (Roche Diagnostics). For all assays, intrarun precision coefficient of variation was <4%.

BMDMs were prepared as described (14, 23) from adult male C57BL/6 mice. Briefly, we differentiated whole bone marrow for 7 d at 37°C, 5% CO₂ in DMEM/F12+Glutamax (Life Technologies) medium with 10% FCS and 25 μg/ml recombinant murine (rm) M-CSF (Miltenyi Biotec) under nonadherent conditions using Ultra-Low attachment flasks (Corning). This process routinely yielded a macrophage population of >90% purity as assessed by flow cytometry for CD11b. ApoTs were prepared as previously described (24). Briefly, thymuses were removed from C57BL/6 mice, aged 3–5 wk, homogenized in RPMI 1640 medium, and incubated with hydrocortisone (1 μM; Sigma-Aldrich) and 1% FCS at 37°C, 5% CO₂ for 16 h. This process routinely yielded a population of dead thymocytes with over 80% trypan blue positive. ApoTs were labeled using CMTMR (Invitrogen) as described previously (25). Briefly, macrophages were challenged with apoptotic cells for 7, 15, 30 min, 1 and 2 h at a 1:5 ratio at 37 or 4°C. Cells were washed and phagocytosis verified by flow cytometry after staining with anti-CD11b BV650 (clone M1/70; eBioscience) and anti–Ly-6C V450 (clone HK1.4; eBioscience). When the assay was performed in a 96-well plate (Ultra-Low attachment; Corning Costar) cytochalasin D 10 μM (Sigma-Aldrich) was used as a negative control. The phagocytosis was calculated as percentage of CD11b⁺CMTMR⁺ cells at 37°C minus the percentage of CD11b⁺CMTMR⁺ cells at 4°C. The percentage of Ly6C⁺ cells was calculated in the gate of CD11b⁺CMTMR⁺ cells at 37°C. Mean fluorescence intensity for CMTMR was calculated on the same gate in the same conditions. Data were acquired on an LSRII Fortessa (BD Biosciences) when the assay was performed in tubes. Data were acquired and analyzed on a NovoCyte 3000 (Acea Biosciences) when the assay was performed on a 96 well-plate.

BMDMs were plated (1 × 10⁵ per well) in 96-well CellCarrier microplates (PerkinElmer) overnight before stimulation with appropriate cytokines or blockers (see 2Materials and Methods). Before imaging, BMDMs were stained with NucBlue live cell stain (Thermo Fisher Scientific) and CellMask Deep Red (Thermo Fisher Scientific) plasma membrane stain according to the manufacturer’s instructions. Plates were transferred to Operetta high-content imaging system (PerkinElmer) and allowed to equilibrate at 37°C and 5% CO₂. Phagocytosis was initiated by the addition of pHrodo green zymosan bioparticles (Thermo Fisher Scientific) to the wells. Fluorescent images were taken in the DAPI channel, 488, and 647 nm before, and at 5 min intervals after the addition of bioparticles for a maximum of 150 min. Images were quantified on Columbus image analysis software (PerkinElmer). Macrophages positive for phagocytosis were classified based on a fluorescence intensity (488 nm) >500 and expressed as a fraction of all live cells (NucBlue-positive cells). Mean fraction values were taken from four separate wells per group.

The role of pSTAT3–IL-10–IL-6 pathway in phagocytosis was investigated in vitro in the presence or absence of rmIL-6 (Miltenyi Biotec) at distinct concentrations: 1, 5, and 50 μg/ml. A pSTAT3 (Tyr705) inhibitory peptide or an irrelevant peptide (Merck Millipore) were used at a 30 μM final concentration in vitro. rmIL-10 was used at a final concentration of 10 ng/ml (Miltenyi Biotec) (25). For the Operetta live imaging experiment, BMDMs were treated with pSTAT3 inhibitory peptide either 2 h before the phagocytosis or during the phagocytosis itself.

For immunofluorescence of pSTAT3, BMDMs were fixed at distinct time points after the start of phagocytosis using PFA 4% for 15 min. Permeabilization was performed using TritonX100 (Sigma-Aldrich) for 15 min. To minimize nonspecific binding, we added protein block solution (SpringBio) for 30 min. Primary Abs used were: Phospho-Stat3 (Tyr705) (clone D3A7) XP rabbit mAb #9145 1 μg/ml (Cell Signaling Technology); GPNMB (K-16) goat pAb #Sc-47006 1 μg/ml (Santa-Cruz). Primary Abs were incubated for 3 h at room temperature. Secondary Abs were AlexaFluor488 and AlexaFluor555, respectively. Secondary Abs were incubated in the dark for 1 h at room temperature. BMDMs were then mounted on a slide using an aqueous mounting medium (Fluoromount-G; SouthernBiotech) and imaged using a Leica SP5 confocal microscope. AlexaFluor488 and 555 were detected using band paths of 495–540 and 561–682 nm for 488 and 543 nm lasers respectively.

The role of pSTAT3–IL-6 pathways in phagocytosis was investigated in vivo by injecting 50 μg per mouse of rmIL-6 (Miltenyi Biotec) i.p. IL-6 was injected together with CCl₄ in the single CCl₄ acute liver damage model. IL-6 was injected either with or 2 h after the administration of APAP. In the chronic model of CCl₄ intoxication IL-6 is injected at the same time of CCl₄ suspension to investigate the effect on the macrophage phenotype in Gpnmb⁺ and Gpnmb⁻ mice. A STAT3 inhibitory peptide or an irrelevant peptide (Merck Millipore) were used at a 30 μM final concentration following the same treatment regimen in the acute damage model by single CCl₄ injection and in the model of APAP intoxication.

For whole liver, the caudate lobe was snap-frozen in liquid nitrogen and stored at −80°C. RNA was extracted using QIAshredder columns and RNeasy mini columns (Qiagen) according to the manufacturer’s protocol, followed by quantification using the Nanodrop Spectrophotometer (Thermo Fisher Scientific); 0.5 μg RNA was reverse transcribed using SuperScript III (Invitrogen) according to the manufacturer’s protocol. Gene expression was calculated using the ΔΔ cycle threshold (CT) method relative to housekeeping gene β-actin and Gapdh. For the in vitro phagocytosis assay, RNA was extracted using the QIAshredder columns and RNeasy mini columns (Qiagen), and 100 ng RNA were reverse transcribed using QuantiTect Reverse Transcription Kit (Qiagen) according to the manufacturer’s protocol. cDNA was then diluted to 1:10 with RNase Free water, prior to quantitative PCR (qPCR) analysis. The following QuantiTect Primer Assays (Qiagen) were purchased: Tgf-β, Il-10, Il-10, Il-6, Hmgb1, Sdf1/Cxcl12, Mcp1/Ccl2, Socs3, col1a2 and col3a1, Mmp2, Mmp7, a-SMA. Genes were analyzed using the Quantifast SYBR Green PCR Kit (Qiagen) on an ABI 7500 Fast Real-Time System or a Roche LightCycler480 according to the manufacturer’s instructions.

Liver tissue was harvested and fixed overnight in 10% neutral buffered formalin or methacarn solution followed by paraffin embedding. Tissue sections were deparaffinized with xylene and rehydrated using alcohol (100, 75, and 65%). They were then subjected to Ag retrieval by pretreating in a microwave oven with TRIS EDTA (pH 9) for CD3 and Ki67 Ag detection. Ag retrieval for a-SMA, collagen 1, and 3 was performed by preheating sections in a microwave oven with citric acid. The following primary Abs and conditions were used: CD3 (rabbit polyclonal; Abcam), Ki67 (rabbit polyclonal; Dako, discontinued), collagen 1 (131008; 1:100 dilution; formalin-fixed; Ag retrieval; SouthernBiotech), collagen 3 (131001; 1:100 dilution; formalin-fixed; Ag retrieval; SouthernBiotech), α-smooth muscle actin (α-SMA; clone1A4; 1:4000 dilution; formalin-fixed; Ag retrieval; Sigma-Aldrich). Endogenous peroxidase activity was inhibited with 3% hydrogen peroxide, and protein block solution (Dako). To minimize nonspecific binding, normal goat serum was added. Appropriate biotinylated secondary Abs were used. Immunostaining was developed using 3,3′-diaminobenzidine (Dako), and counterstaining with Harris’s hematoxylin. Positive cells per area were measured from 20 random fields at ×20 magnification. Sections were photographed using a Nikon Eclipse E600 microscope and NIS-Elements D3.1 Software. H&E staining was performed according to standard protocols. Morphometric pixel analysis to quantify histological staining was performed. For necrosis quantification, H&E-stained sections were scanned to create a single image with Dotslide VS-ASW software (Olympus) using a motorized stage and an Olympus BX51 microscope, acquiring images using an Olympus PlanApo 2× lens and Olympus XC10 camera. Images were analyzed using the Trainable WEKA Segmentation plugin in FIJI. A separate classifier identifying necrotic and viable tissue was determined and applied to all tissue in each image.

Human monocyte–derived macrophages (hMDMs) were differentiated from cryopreserved CD14 monocytes essentially as described previously (26). Briefly, cryopreserved stocks were thawed rapidly and diluted in IMEM supplemented with 10% FBS (v/v), 2 mM glutamine, penicillin/streptomycin (500 U/ml, 500 μg/ml), and 100 ng/ml human recombinant CSF1 (Miltenyi) at 2 × 10⁶ cells per ml in Ultra-Low attachment flasks (Corning). Cells were differentiated toward macrophages for 7 d with a 10% media change every second day containing 1 μg CSF1. After 7 d, hMDMs were harvested, counted, and plated on CellCarrier plates for in vitro phagocytosis assays using the Operetta instrument for live imaging (see above).

Adult zebrafish (Danio rerio) strain TgBAC(csf1ra:GAL4-VP16;UAS:mCherry) (27) were maintained at 28°C and crossed as previously described (28). Embryos were maintained at 28°C and staged according to standard protocols (28).

Sterile tail fin amputation was performed as previously described (29). Briefly, zebrafish embryos at 2 d postfertilization were anesthetized in 0.3% Danieau’s solution containing 0.1 mg/ml tricaine (ethyl 3-aminobenzoate; Sigma-Aldrich) and tail fins were cut off from the end of the notochord using a scalpel. Regeneration was monitored at 24, 48, and 72 h after wounding. Photographs of the regenerating fins were taken at 40× or 50× using a Leica M205 stereomicroscope.

Next, 5,15-diphenyl-21H,23H-porphine, 5,15-diphenylporphyrin, 5,15-diphenyl-porphine was dissolved in DMSO (stock 10 mM) and diluted to a concentration of 400 μM and 2 mM in 0.3% Danieau’s solution. Larvae were treated with compounds at their final concentration immediately after amputation, and media were changed daily with fresh 0.3% Danieau’s solution containing the compound.

All data are expressed as mean ± SD. The number of replicates is indicated in each figure and each replicate represents a biological rather than an experimental replicate. Data are analyzed and graphs are generated with GraphPad Prism version 5 or 6 (GraphPad Software). Statistic tests have been chosen depending on the biological question behind the experiment. Briefly, we used one- or two-way ANOVA followed by an appropriate posthoc test. The test used is stated in each figure legend. A p value < 0.05 is considered statistically significant.

We have performed a power calculation for the number of mice to use in the studies on the chronic CCl₄ model based on a pilot study on Gpnmb⁺ and Gpnmb⁻ mice on the ALT measure (indicating liver damage) at 48 h after the withdrawal of CCl₄. We have assumed an mu1 of 150 for Gpnmb⁺ and an mu2 of 600 for Gpnmb⁻, with a σ of 150. We have set the desired power at 0.80 assuming a statistical significance at the threshold of 0.05. The power calculation returned n = 3. This is the minimum number of mice used in each experiment.

For the in vivo phagocytosis there were no preliminary data available. We have treated from a minimum of three to a maximum of seven mice per group. The parameter analyzed for the power calculation is the % of phagocytic infiltrating inflammatory macrophages. If we assume an mu1 (Gpnmb⁺) of 80 and an mu2 (Gpnmb⁻) of 20 as derived from our experiment, and a σ of 15, the power calculation for a statistical significance set at 0.05 returns a power of 0.80 with n = 3.

We have used six mice per group in the adoptive transfer experiment. Also in this case no preliminary data were available. If we assume an mu1 of 2 for Gpnmb⁺ and an mu2 of 5 for Gpnmb⁻, as suggested by our experiment, with a σ of 1.7 and a statistical significance at the threshold of 0.05, the power calculation for a statistical significance set at 0.05 returns a power of 0.80 if n = 6 as in our experiment.

For all experiments a two-sided test was considered. All data were tested for normal distribution and equal variance before performing any statistical analysis using Prism v5 or v6. Power calculation has been made using the free online tool available at http://www.stat.ubc.ca.

To investigate pathways regulating the late stages of efferocytosis, we tested BMDMs from Gpnmb⁺ and Gpnmb⁻ mice, which are deficient for the last step of phagocytosis. We analyzed the RNA from Gpnmb⁺ and Gpnmb⁻ BMDMs fed with apoTs (apoT) on a low-density qPCR array for inflammatory cytokines and chemokines. One of the most downregulated genes in Gpnmb⁻ BMDMs is Il-6 (Fig. 1A, 1B, Supplemental Fig. 1A). Gpnmb⁻ BMDMs are able to initiate efferocytosis but internalization of apoT rapidly tails off (Fig. 1C and gating strategy Supplemental Fig. 1B). Il-6 mRNA is increased at 30 and 60 min after addition of apoT to WT BMDMs (Fig. 1D) and IL-6 protein levels are constant for up to 120 min (Fig. 1E, Supplemental Fig. 1C). This suggests that IL-6 is secreted in a cargo digestion-dependent manner. We hypothesized a role for IL-6 in sustaining efficient efferocytosis in macrophages. To this end, we tested, by flow cytometry, the ability of Gpnmb⁺ and Gpnmb⁻ BMDMs to phagocytose apoT in the presence or the absence of increasing concentrations of rmIL-6. rmIL-6 increases the percentage of Gpnmb⁻ phagocytic BMDMs at any concentration tested (2 h duration of efferocytosis, Fig. 1F). To confirm the role of IL-6 in vivo, we induced acute liver damage by a single injection of CCl₄ in Gpnmb⁺ and Gpnmb⁻ mice and we treated them with rmIL-6 or vehicle 6 h prior to culling (Fig. 1G). We digested the livers and analyzed the percentage of infiltrating macrophages by flow cytometry (CD45⁺Lin⁻CD11b^highF4/80^low). In this gate, we analyzed the percentage of phagocytic (PKH26PCL⁺), inflammatory (Ly6C^high), or restorative (Ly6C^low) macrophages. Whereas rmIL-6 treatment did not affect the overall percentage of infiltrating macrophages (Fig. 1H and gating strategy in Supplemental Fig. 1E), it normalized to a WT pattern the phenotype of Gpnmb⁻–infiltrating inflammatory macrophages in vivo. It had less effect on restorative macrophages (Ly6C^low), which infiltrate as inflammatory macrophages before undergoing a phenotypic switch (14) (Fig. 1I and representative plots Fig. 1J).

We then analyzed possible pathways upstream of IL-6. Using two inhibitors of NF-κB, we demonstrated that there is no role for this transcription factor in IL-6 regulation of efferocytosis (Supplemental Fig. 1D). IL-6 is one of the cytokine pathways known to induce hepatocyte proliferation in models of hepatocellular carcinoma (HCC) via activation of pSTAT3. In these studies, IL-6 is identified as a single factor coregulating key aspects of macrophage-mediated tissue remodelling and phenotype with epithelial regeneration (30, 31). We tested the hypothesis that this pathway sustains macrophage efferocytosis in a cell-autonomous manner in BMDMs. pSTAT3 blockade resulted in impaired phagocytosis in BMDMs 2 h after the start of phagocytosis (Fig. 2A, 2B) as assessed by flow cytometry. We confirmed the results using live imaging: blocking pSTAT3 impairs phagocytosis of zymosan A–coated, pH sensitive (pHrodo) beads in WT BMDMs and in hMDMs alike (Fig. 2C, 2D), showing the conserved nature of this pathway. In WT BMDMs the impairment is evident as soon as 30–35 min after the start of phagocytosis, and is more complete than blockade of the acidification of the phagosome with bafilomycin (Fig. 2C and Supplemental Videos 1, 2). The pHrodo beads emit fluorescence when in acidic compartments and empty vacuoles are observed in the cytoplasm of BMDMs in the presence of the pSTAT3-blocking peptide (Supplemental Video 3). A similar but less dramatic effect is observed if pSTAT3 is blocked in BMDMs prior to the induction of phagocytosis, indicating that the prophagocytic activity relies mainly on de novo phosphorylated STAT3 (Supplemental Fig. 2A). The blockade of pSTAT3 decreases phagocytosis of infiltrating inflammatory macrophages (Ly6C^high) in vivo in a model of acute liver damage induced by CCl₄ injection in which a phagocytic cell tracker is injected (PKH26PCL, for phagocytic cell labeling) (Fig. 2E). The percentage of infiltrating inflammatory macrophages positive for PKH26PCL (i.e., performing phagocytosis) is lower in livers from Gpnmb⁺ mice treated with the pSTAT3-blocking peptide as compared with Gpnmb⁺ mice treated with an irrelevant peptide, and similar to livers from Gpnmb⁻ mice (Fig. 2F and representative plot in Fig. 2G). A similar trend is observed in infiltrating restorative macrophages (Fig. 2F). No effect of pSTAT3 inhibition is observed on the overall percentage of resident macrophages (CD45⁺Ly6G⁻CD11b^lowF4/80^high), and infiltrating macrophages in Gpnmb⁻ mice (Fig. 2H, Supplemental Fig. 2B, respectively, and gating strategy in Supplemental Fig. 1E).

To confirm that IL-6 is downstream of pSTAT3, we treated BMDMs with the blocking peptide for pSTAT3, and rescued their phagocytosis using rmIL-6. Live imaging results show that rmIL-6 is able to rescue the BMDM phenotype at 2 h of phagocytosis (Fig. 3A, 3B). The result is confirmed by flow cytometry (Fig. 3C). Two hours after the start of efferocytosis Il-6 mRNA is reduced in the presence of the pSTAT3 blocking peptide (Fig. 3D); consistently, IL-6 protein shows a reduced trend in the supernatant of phagocytosing BMDMs when pSTAT3 is blocked (Fig. 3E). Therefore, pSTAT3 and IL-6 appear to be in the same pathway, effective in sustaining phagocytosis, with IL-6 controlled at the transcriptional level by pSTAT3 when macrophages are exposed to apoptotic cells for more than 60 min. We then explored the dynamics of pSTAT3 activation upon phagocytosis using flow cytometry, to interrogate the phagocytic fraction of BMDMs (CD11b⁺CMTMR⁺). pSTAT3 is activated as early as 7 min after the start of phagocytosis and its activation is sustained at any time point analyzed (Fig. 3F, Supplemental Fig. 2D). Gpnmb⁻ BMDMs can internalize the cargo but they cannot digest it. Consistent with a role for the pSTAT3–IL-6 axis in sustaining efferocytosis at the cargo digestion stage, pSTAT3 can be activated in Gpnmb⁻ BMDMs but its activation fails to be sustained at later time points (Fig. 3G, Supplemental Fig. 2E) and pSTAT3 intracellular localization is disrupted in Gpnmb⁻ BMDMs (Supplemental Fig. 2G). However, analysis of pSTAT3 activation during phagocytosis following rmIL-6 treatment showed no significant difference (data not shown).

Later time points of phagocytosis are associated with a cargo digestion-dependent activation of pSTAT3. However, the consequence of the early phosphorylation of STAT3 is thus far unexplored. A target molecule of pSTAT3 is IL-10, and links between IL-10 and phagocytosis have been previously reported (32, 33). We hypothesized that IL-10 could be downstream of pSTAT3 early activation and have a prophagocytic role. To this end, we treated Gpnmb⁺ and Gpnmb⁻ BMDMs with rmIL-10 and measured their phagocytic ability. Gpnmb⁺ BMDMs showed increased phagocytosis at 15 min and Gpnmb⁻ BMDMs demonstrated a similar trend (Fig. 4A, 4B, Supplemental Fig. 2F). IL-10 protein increased at 15 min in the supernatants of fed BMDMs (Fig. 4C), pointing to a possible early nontranscriptional regulation of IL-10. To confirm a role for pSTAT3 in the regulation of IL-10 secretion early after the start of efferocytosis, we blocked pSTAT3 and measured IL-10 protein in the supernatants of fed WT BMDMs. IL-10 protein dramatically decreases at 15 and 30 min (Fig. 4D). Il-10 transcription is not induced in BMDMs performing efferocytosis until 30–60 min (Fig. 4E), thereby suggesting a nontranscriptional control of IL-10 secretion by pSTAT3. Consistently, the levels of Il-10 mRNA are substantially stable when we treat BMDMs with the blocking peptide for pSTAT3 during efferocytosis (Fig. 4F). Thus, pSTAT3 appears to regulate IL-10 release at early stages of efferocytosis in a transcriptionally independent manner. Consistent with a role of IL-10 at early stages of phagocytosis, we did not observe any difference in IL-10 levels in the supernatants of Gpnmb⁺ and Gpnmb⁻ macrophages after overnight efferocytosis (Supplemental Fig. 2H). SOCS3 may play a role in macrophage phagocytosis and polarization. Moreover, SOCS3 is known to support IL-6 transcription (34, 35). Consistent with a role of Socs3 in IL-6 production and signaling, we found that Socs3 transcription is controlled by pSTAT3 activation and cargo digestion at 1 h after the start of phagocytosis in BMDMs (Fig. 4G, 4H). To show that the role of pSTAT3 is conserved in tissue repair and remodelling across species, we used the tail fin injury model in D. rerio (zebrafish) embryos. D. rerio embryos were harvested 48 h postfertilization and we performed the tail injury at the level of the notochord. First, we replicated recent data suggesting that blocking the acidification of intracellular compartments delays tail fin regeneration, a defect similar to that of Gpnmb⁻ mice (18, 36). To this end, embryos were left to recover from tail fin injury in the presence or absence of bafilomycin to block the acidification of intracellular compartments; length and area of the regenerating fin were recorded at 24, 48, and 72 h postdamage. Treated embryos recapitulated the phenotype of the Gpnmb⁻ mice with a lower regeneration of the tail fin at 72 h postinjury (data not shown). To test the hypothesis that tail fin remodelling and regeneration is pSTAT3 dependent, we repeated the same experiment monitoring the regrowth of the tail fin in the presence or absence of a small molecule inhibiting pSTAT3, 5,15-diphenyl-21H,23H-porphine, 5,15-diphenylporphyrin, 5,15-diphenyl-porphine (Supplemental Fig. 4F). Reduced tail fin remodelling and regeneration was evident at 24 h after the injury induction. By 48 and 72 h after injury the tail fin remodelling and regeneration was dramatically impaired in terms of both length and area (Supplemental Fig. 4G, 4H). Although we cannot prove that the blockade to tail fin regeneration is macrophage mediated, our data provide a link between pSTAT3 and tissue repair in both zebrafish and mice.

Sterile liver injury has proven a useful model to define the interplay of parenchymal cells and inflammation in the mammalian wound-healing response. Therefore, we proceeded to investigate the potential role of cargo digestion during efferocytosis as a regulator of tissue damage and macrophage phenotype in a model of chronic sterile liver injury. We treated Gpnmb⁺ and Gpnmb⁻ mice with CCl₄ for 6 wk to induce iterative liver parenchymal sterile necrosis and inflammation with resulting liver fibrosis (Fig. 5A). At the suspension of CCl₄ administration we performed a time-course analysis of circulating pALT and pAST as markers of liver damage, together with pALP, albumin, urea, and creatinine (Fig. 5B, 5C). The liver/body weight ratio was lower in Gpnmb⁻ mice at 72 h, whereas the two groups of mice show a comparable weight increase during the 6 wk of treatment (Supplemental Fig. 3A, 3B). pALT and pAST are specifically increased in Gpnmb⁻ mice at 48 h after the suspension of the CCl₄ and the trend is maintained at 72 h. At 1 wk, the levels of ALT and AST are comparable in the two groups of mice. pALP, albumin, urea, and creatinine are unchanged in either genotype at any time point analyzed (Fig. 5B, 5C). We then analyzed the infiltrating macrophages in the damaged liver and the circulating monocytes by flow cytometry (gating strategy in Supplemental Fig. 3C). Failure of phagocytic cargo processing does not affect the general percentage of infiltrating macrophages (live CD45⁺Ly6G⁻CD11b^highF4/80^low) at the basal level and at any time point studied (Fig. 5D, 5H). We then analyzed the percentage of restorative macrophages [which we and others have shown can be identified as Ly6C^low (14)]. In Gpnmb⁻ livers, the percentage of Ly6C^low infiltrating macrophages was reduced relative to Gpnmb⁺ controls at any time point analyzed (Fig. 5E, 5F), suggesting that phagocytosis is a driver of the switch to a Ly6C^low restorative macrophage phenotype. Analysis of the expression of liver cytokines and chemokines in the whole liver showed unchanged levels of Mcp1/Ccl2, Cxcl12, and Hmgb1 (Supplemental Fig. 3D). Consistent with the lack of difference in the overall infiltration of macrophages, T lymphocyte infiltrate was comparable in the two groups of mice at all time points analyzed and at basal level (Fig. 5G). Circulating granulocytes, monocytes, and, particularly, classic monocytes were similar in the two groups of mice (Fig. 5I). Overall, the impairment in phagocytic cargo processing caused increased liver damage and prevented macrophage phenotype switching during regeneration. If IL-6 has a prophagocytic role relevant for the clearance of damaged cells, then we would expect an increase in damage in the livers of IL-6 KO mice with iterative sterile injury (Fig. 5J). Supporting this model and directly reproducing the phenotype observe in Gpnmb⁻ mice, IL-6 KO mice showed higher levels of circulating ALT and AST at 48 h after CCl₄ withdrawal, without showing any difference in the level of pALP (Fig. 5K) or in their liver/body weight ratio (Supplemental Fig. 4A).

We then verified the impact of impaired phagocytic cargo processing on liver pathology. At 1 wk of recovery, there is no difference in fibrosis between the two groups of mice (Fig. 6A, 6B). Gpnmb⁻ BMDMs showed similar level of Tgf-β when compared with Gpnmb⁺ BMDMs after overnight phagocytosis (Fig. 6C). Further, the lack of difference in liver fibrosis between the two groups of mice was confirmed by collagen 1, Tgf-β, Mmp-2, and Mmp-7 mRNA analysis (Fig. 6D, 6F, 6I, 6L). Col3a1 mRNA was increased at 48 h (Fig. 6G), but no differences between the two groups of mice were observed in the immunohistochemical analysis (Fig. 6H). However, Gpnmb⁻ livers show increased α-SMA levels at 48 h of recovery (Fig. 6J, 6K). The similar levels of liver fibrosis in Gpnmb⁻ mice may result in part from the increased proliferation of parenchymal cells at 48 h of recovery (Fig. 6M, 6N).

We collected further evidence for a correlation between phagocytosis, IL-6, and damage resolution in a clinically relevant model of sterile acute liver damage by APAP intoxication (Fig. 7A), which is characterized by extensive parenchymal cell death and inflammation (37, 38). Gpnmb⁻ mice have a higher level of pALT at 8 h postinjury (Fig. 7B); consistently, they showed a more extended necrotic area than their Gpnmb⁺ counterparts as assessed by image analysis (Fig. 7C, 7D and Supplemental Fig. 3E). We then dosed C57BL/6 WT mice with APAP and treated them with rmIL-6 at the same time as, or 2 h after, the APAP injection. To track phagocytic cells, we injected PKH26PCL into some of the mice, which were culled at 8 h after APAP dosing to quantify the effect of rmIL-6 treatment on the phagocytic ability of infiltrating macrophages (Fig. 7E). We did not observe a significant effect of IL-6 treatment on the necrotic area, the pALT or the liver/body weight ratio (Fig. 7F, 7G, Supplemental Fig. 4B). Importantly, the number of phagocytic macrophages per gram of liver was increased in rmIL-6–injected mice (Fig. 7H, 7I). The treatment was efficacious in increasing macrophage phagocytosis without affecting the percentage of infiltrating macrophages and granulocytes in the damaged liver (Fig. 7J, 7K). The local effect of the systemic treatment with rmIL-6 is confirmed by a lack of effect on the phagocytic ability of circulating monocytes (Fig. 7L).

To test the hypothesis that the prophagocytic positive feedback loop triggered by the STAT3–IL-6 axis drives the conversion of macrophage phenotype, and thereby provides a link between tissue damage and tissue repair, we administered rmIL-6 to Gpnmb⁺ and Gpnmb⁻ mice after induction of chronic liver injury by CCl₄. Mice were culled at 48 and 72 h after CCl₄ withdrawal (Fig. 8A). rmIL-6 restored the ability of Gpnmb⁻ mice to convert the inflammatory Ly6C^high macrophages into restorative Ly6C^low macrophages within 72 h of administration (Fig. 8B, 8C). Consistent with a reprogramming action of phagocytosis, WT BMDMs fed with CMTMR-labeled apoTs showed a sharp decrease of the percentage of Ly6C^high events; i.e., bona fide inflammatory BMDMs were reduced just as the percentage of phagocytic BMDMs increased in the population (Fig. 8D, 8E). Independent evidence was sought by treating APAP-dosed mice with rmIL-6 at distinct time points (Fig. 7E). The number of Ly6C^high infiltrating macrophages (i.e., infiltrating inflammatory macrophages) was lower in mice treated with rmIL-6 (Fig. 8F), which also showed a higher number of phagocytic macrophages infiltrating the liver (Fig. 7H). To demonstrate that phagocytosis triggers a macrophage-phenotypic switch, we induced acute damage by a single injection of CCl₄ in C57BL/6 mice. After 3 h, we adoptively transferred CFSE⁺Gpnmb⁺ and CMTMR⁺Gpnmb⁻ BMDMs separately, or mixed in 1:1 ratio. Then 5 h later mice are culled, the liver digested, and cells analyzed (Fig. 9A, gating strategy in Supplemental Fig. 4C). Macrophage conversion to a restorative phenotype was monitored using the level of Ly6C expression. The percentage of Ly6C^high adoptively transferred macrophages was calculated in Gpnmb⁺ and Gpnmb⁻ BMDMs before and after transplant. Both genotypes showed a lower posttransfer percentage of Ly6^high transferred macrophages as compared with pretransfer (Fig. 9B). However, adoptively transferred Gpnmb⁻ macrophages showed a consistently higher percentage of Ly6C^high cells versus their Gpnmb⁺ counterparts cotransferred in the same mouse and therefore in an identical microenvironment (Fig. 9B–D). The higher percentage of Ly6C^high cells in Gpnmb⁻ macrophages was also confirmed when the data from the mice receiving only Gpnmb⁺ or Gpnmb⁻ BMDMs were included in the analysis (Fig. 9E, 9F). The transfer of Gpnmb⁻ BMDMs reduced the hepatic recruitment of the endogenous macrophages: host livers showing a lower number of both Ly6C^high and Ly6C^low infiltrating macrophages and a contraction of the Ly6C^low population (Fig. 9H). The difference observed was not due to a difference in the viability of extracted cells (Fig. 9G, Supplemental Fig. 4D, 4E).

Our data suggest that the processing of apoptotic cells activates a STAT3–IL-10–IL-6 autocrine-paracrine loop, enabling macrophages to maintain their scavenging ability. Furthermore, a failure of phagocytic cargo processing results in the reduced uptake of debris (predominantly in proinflammatory macrophages), prolonged liver damage, enhanced hepatocyte proliferation, and a reduced expression of IL-10. These data suggest that macrophage ingestion of apoptotic cell and IL-6 production are important for debris clearance and the phenotypic switch to a restorative phenotype. Apoptotic cells are necessary to trigger an IL-4/IL-13–dependent tissue repair response in models of helminth infection (39). In this study, we provide data linking an immediate event in proinflammatory macrophages to the programming of tissue remodelling and regeneration in the context of sterile injury. Phagocytosis of apoptotic bodies (efferocytosis) may require the presence of autophagy-related molecules for the assembly of the phagosome, a process named LC3-associated phagocytosis (40). During LC3-associated phagocytosis some of the autophagy machinery, including LC3, is recruited to pathogen, apoptotic, and necrotic cell-containing phagosomes; as a result, optimal degradation of the phagocytosed cargo is achieved (41, 42). Gpnmb is an interactor of LC3 (18), making Gpnmb⁻ mice a useful model to study the effect of an incomplete phagocytic process of apoptotic cells on sterile injury and repair models.

In our model, STAT3, IL-10, and IL-6 are in the same pathway and provide mechanistic insight on previous observations that correlated IL-10 and phagocytosis in in vitro models (32, 33, 43, 44). Our work provides a mechanistic explanation to recent data regarding the restorative role of Gpnmb⁺ macrophages in acute liver (45) and kidney injuries (18), and in DSS-induced colitis (46); and the proresolution role of Gpnmb⁺ macrophages in a model of peritonitis (47). Growing evidence correlates pSTAT3 activation with tumor cell proliferation, genomic instability, and migration (48–53). In addition, recent studies establish a link between pSTAT3 activation and immunomodulation in tumors, via infiltrating monocytes switching phenotype to tumor-associated macrophages (54–57). Moreover, the IL-10–IL-6–pSTAT3 pathway has been reported as having a role in the inflammatory process in inflamed adipose tissue from obese patients (58). Our data suggest pSTAT3 acts as a driver of inflammatory macrophage phagocytosis and phenotypic conversion. Although the switch to an anti-inflammatory, pro-remodeling phenotype is detrimental in tumor models (11, 59), it is required in physiological tissue repair (1, 9). The completion of a correct phagocytic process leads macrophages to switch their inflammatory phenotype into a restorative one, evidencing that phagocytosis is a major process through which the initial damage phase initiates the repair of the tissue. Our in vitro data and in vivo adoptive transfer show this process happens soon after the induction of damage. The identification of phagocytosis as a key element mediating macrophage phenotype conversion is of therapeutic relevance for a number of diseases. Means to increase restorative-like macrophage phenotype have proved beneficial to block CNS inflammation (60), systemic lupus erythematosus (61), and to contribute to hepatic progenitor cell specification in models of liver disease (16, 62). STAT3-IL-6 is a pro-proliferative signal in HCC development. Tumor-associated macrophages in HCC have a phenotype similar to restorative macrophages (63). Macrophages forced to stay in an inflammatory phenotype downregulate Stat3 and have a higher anti-HCC activity (64). Our data linking a STAT3–IL-10–IL-6 positive feedback loop with phagocytosis enhancement and a macrophage phenotypic switch may also explain recent data showing that IL-6–producing macrophages, polarized with rIL-4, block neuroinflammation in vivo (65). The possible link between STAT3–IL-10–IL-6 and the control of hepatocyte proliferation is worthy of further investigation. Macrophages link tissue necrosis and repair in many diseases; explaining how to program their regenerative response via the control of their ability to scavenge dead cells will be important for future therapeutic targeting in multiple clinical settings, which may include acute liver injury and chronic liver fibrosis.

We thank Dr. William Ramsay and Dr. Shonna Johnston at the Flow Cytometry Facility of the Queen’s Medical Research Institute of the University of Edinburgh for the technical assistance during the first part of the project. We thank Dr. Fiona Rossi and Dr. Clare Cryer at the Flow Cytometry Facility of the Medical Research Council Centre for Regenerative Medicine of the University of Edinburgh for technical assistance during the second part of the project. We thank the Shared University Research Facilities at the University of Edinburgh for assistance with histology and qPCR. We thank Dr. Bertrand Vernay at the Imaging Facility of the Medical Research Council Centre for Regenerative Medicine of the University of Edinburgh for support with the confocal imaging. We thank the technicians in the animal units of Little France and of the Tissue Culture Unit of the Medical Research Council Centre for Regenerative Medicine for continuous help and support throughout the project. In particular, we thank Laraine Wells, William Mungall, Lynn, and Theresa O’Connor.

The authors have no financial conflicts of interest.