M2 Kupffer cells promote M1 Kupffer cell apoptosis: A protective mechanism against alcoholic and nonalcoholic fatty liver disease


  • Potential conflict of interest: Nothing to report.


Alcoholic and nonalcoholic fatty liver disease (ALD and NAFLD) are the predominant causes of liver-related mortality in Western countries. We have shown that limiting classical (M1) Kupffer cell (KC) polarization reduces alcohol-induced liver injury. Herein, we investigated whether favoring alternatively activated M2 KCs may protect against ALD and NAFLD. Ongoing alcohol drinkers and morbidly obese patients, with minimal hepatic injury and steatosis, displayed higher hepatic expression of M2 genes, as compared to patients with more severe liver lesions; individuals with limited liver lesions showed negligible hepatocyte apoptosis but significant macrophage apoptosis. Experiments in mouse models of ALD or NAFLD further showed that BALB/c or resveratrol-treated mice fed alcohol or a high-fat diet displayed preponderant M2 KC polarization, M1 KC apoptosis, and resistance to hepatocyte steatosis and apoptosis, as compared to control C57BL6/J mice. In vitro experiments in isolated KC, peritoneal, and Raw264.7 macrophages demonstrated that M1 macrophage apoptosis was promoted by conditioned medium from macrophages polarized into an M2 phenotype by either interleukin (IL)4, resveratrol, or adiponectin. Mechanistically, IL10 released from M2 cells promoted M1 death, and anti-IL10 antibodies blunted the proapoptic effects of M2-conditioned media. IL10 secreted by M2 KCs promoted selective M1 death by a mechanism involving activation of arginase in high inducible nitric oxide synthase-expressing M1 KCs. In alcohol-exposed mice, neutralization of IL10 impaired M1 apoptosis. Conclusion: These data uncover a novel mechanism regulating the M1/M2 balance that relies on apoptotic effects of M2 KCs towards their M1 counterparts. They suggest that promoting M2-induced M1 KC apoptosis might prove a relevant strategy to limit alcohol- and high fat-induced inflammation and hepatocyte injury. (Hepatology 2014;58:130–142)


alcoholic liver disease


Kupffer cell


nonalcoholic liver disease.

The functional plasticity of macrophages is driven by their immunological environment that can shape their properties through a wide spectrum of phenotypes, among which classical (M1) or alternative (M2) represent the extreme states.[1, 2] M1-polarized macrophages play a key role in a variety of chronic inflammatory diseases, such as atherosclerosis,[3] inflammatory bowel disease,[4] or insulin resistance associated with obesity.[5] Inflammation driven by M1 macrophages is counterbalanced by alternatively polarized M2 macrophages that promote resolution of inflammation and tissue repair.[1, 2] Beneficial properties of alternative M2 macrophages are reported in several inflammatory disorders, including insulin resistance,[6, 7] atherosclerosis,[3] muscle repair,[8] infectious colitis,[4] and chronic glomerulopathies.[9] Altogether, dysregulation of the M1/M2 phenotypic balance is emerging as a central mechanism governing the pathogenesis of chronic inflammatory diseases, suggesting that strategies restraining M1 macrophage polarization and/or favoring the M2 macrophage phenotype may protect against exacerbated inflammation and thus limit tissue injury.

Alcoholic and nonalcoholic fatty liver disease (ALD and NAFLD) are leading causes of liver-related morbidity and mortality in Western countries.[10, 11] Clinical findings and experimental data have demonstrated that activation of Kupffer cells (KCs) is a central event in the initiation of liver injury.[11-14] The resulting exacerbated release of M1 Kupffer cell-derived mediators contributes to the pathogenesis of several liver lesions, namely, hepatocyte steatosis and apoptosis, inflammatory cell recruitment, and activation of fibrogenesis.[14-17] We have previously shown that limiting M1 KC polarization reduces ALD progression.[6, 7, 14] Here we investigated the impact of M2 KC polarization on alcohol- and high fat-induced liver injury and uncover a novel mechanism limiting M1 KC function, which relies on a proapoptotic effect of M2 KCs for their M1 counterparts.

Materials and Methods

Human Study

Study Population

All subjects gave their informed written consent to participate in this research study according to French legislation regarding Ethic and Human Research (Huriet-Serusclat law, the “Comité Consultatif de Protection des Personnes dans la Recherche Biomedicale de Nice” approved this study, Nos. 03.613 and 03.017).

Ongoing Alcohol Drinker Patients

The M2 signature was investigated in liver biopsies from 15 heavy alcohol drinkers with similar age and daily alcohol intakes, and limited fibrosis, as defined by a fibrosis stage <2 according to the modified METAVIR scoring system (Table 1). Patients were part of a previously characterized cohort of 109 consecutive alcohol drinkers.[18] Patients were divided into two groups according to serum transaminase levels: group 1 with minimally increased transaminases (10 patients with aspartate aminotransferase [AST], < 2 ULN [(upper limit of normal] and normal alanine aminotransferase [ALT]), and group 2 with elevated transaminases (five patients with AST >2× ULN and ALT >ULN). All patients had ongoing alcohol intoxication until admission. Ultrasound-guided liver biopsy was performed after a mean period of 4 ± 1.9 days of alcohol abstinence, with no difference in duration of abstinence between groups (group 1; 4.5 ± 1.1 days; group 2: 3.5 ± 1.3 days; P = 0.2696).

Table 1. Characteristics of Patients
 Group 1: Minimally Increased Transaminases (AST < 2x ULN and ALT < 1x ULN)Group 2: Elevated Transaminases (AST > 2x ULN and ALT > 1x ULN)P
  1. Values are expressed as median (25th-75th percentile). ULN: upper limit of normal. S0: <5% steatotic hepatocytes; S1: 5-30% steatotic hepatocytes; S2: 30-60% steatotic hepatocytes.

Heavy alcohol drinkers with minimally increased (n = 10) and elevated (n = 5) serum transaminase values
N (female/male)10 (2/8)5 (1/4) 
Age (year)44 (37-52)46 (44-51)0.53
duration of alcohol intake (year)12 (7-27)11 (7-25)0.9
alcohol (g/day)165 (120-265)160 (95-205)0.58
AST (ULN: 37 IU/L)39 (28-47)153 (99-209)0.0027
ALT (ULN: 41 IU/L)33 (12-39)81 (62-190)0.0027
γGT (ULN: 65 IU/L)85 (69-116)451 (199-1357)0.0047
Serum caspase-generated cK18 fragment (IU/L)289 (252-387)681 (551-851)0.0013
Steatosis N (S1/S2)10 (8/2)5 (1/4)0.047
Fibrosis< F2 (modified METAVIR)105 
 Group S0:Group S2:P
Obese patients with minimal (S0 <5 %, n = 7) and elevated (S2 30-60%, n = 6) Steatosis
N (female/male)7 (6/1)6 (5/1) 
Age (year)35 (25-51)37 (29-50)0.63
BMI (kg/m2)43 (42-44)42 (37-46)0.94
HOMA-IR2.2 (1.5-2.1)2.9 (1.3-5.3)0.83
AST (ULN: 37 IU/L)18 (16-20)21 (15-25)0.29
ALT (ULN: 41 IU/L)15 (12-19)27 (19-37)0.0047
γGT (ULN: 65 IU/L)18 (14-23)31 (15-48)0.23
Steatosis N (S0/S2)7 (6/0)6 (0/6)0.0005
N (S0/S2)7 (6/0)6 (0/6)0.0005
Fibrosis< F2 (modified METAVIR)76 

Morbidly Obese Patients

Surgical liver biopsies were obtained from morbidly obese patients (n = 13, Table 1) at the time of bariatric surgery and histological scoring of steatosis evaluated as previously described.[19] Patients were divided into two groups according to steatosis grades, S0 (<5%) and S2 (30%-60%).

Animal Study

Animal procedures were conducted in accordance with French government policies (Comité d'éthique COMETH, Authorization Nos. 10-0048 and 11-0068).

Chronic Exposure to Alcohol

Female C57BL6/J and BALB/c mice were fed for 17 days with a liquid diet adapted from Lieber-De Carli as described.[14]

Early Response to Alcohol

Female C57BL6/J mice were given a single dose of ethanol (5 g/kg body weight, 20% ethanol) or isocaloric maltodextrin by intragastric gavage.

High Fat Diet (HFD) Protocol

Male C57BL6/J mice were fed for 27 weeks with an HFD in which 60% of calories are derived from fat (D12492, Ssniff, Germany), or a normal diet (ND) (11% of calories from fat; 1320, Genestil, France). See the Supporting Materials and Methods for detailed information on experimental designs and methods.


High M2/M1 Ratio Is a Characteristic of Mice Protected Against Alcohol-Induced Liver Lesions

Th2-biased BALB/c mice and C57BL6/J mice were subjected to a Lieber-De-Carli-derived alcohol diet.[14] There was no differences either in daily alcohol intake, serum ethanol level (Supporting Table S1), or alcohol metabolism between the two strains, as attested by similar messenger RNA (mRNA) expression of cytochrome P4502E1, alcohol dehydrogenase, and aldehyde dehydrogenase (not shown). Moreover, livers from both strains of alcohol-fed mice showed negligible signs of inflammatory cell infiltration, with no increase in hepatic expression of F4/80 and CCR2 mRNA (Fig. 1A; Supporting Fig. S1A), in the number of Gr-1-expressing cells (Fig. S1B), and in the density of F4/80-positive cells (Fig. 2A), thus providing a unique opportunity to study the role of resident macrophages.

Figure 1.

Alcohol-fed BALB/c mice display high expression of M2 markers and a resistance towards alcohol-induced M1 polarization, liver injury, and steatosis. (A) Reverse-transcription polymerase chain reaction (RT-PCR) analysis of hepatic M1 and M2 genes was evaluated in BALB/c and C57BL6/J mice fed control diet or alcohol (EtOH) for 10 days (n = 12 mice/CD group and n = 30 mice/ETOH group, two experiments). *P < 0.05 for EtOH versus CD; §P < 0.05 for BALB/c CD versus C57BL6/J CD; &P < 0.05 for BALB/c EtOH versus C57BL6/J EtOH. (B) Alcohol-induced liver steatosis was evaluated by Oil red-O staining (n = 12/group) and quantification of hepatic triglyceride content (n = 6 for CD and n = 15 for EtOH in each group). Results obtained from two experiments. (C) Liver injury was assessed by hepatic cleaved-caspase-3 staining (n = 12 per group) and serum alanine aminotransferase activity (n = 6 for CD and n = 15 for EtOH in each group). Results obtained from two experiments. *P < 0.05 for EtOH versus CD. (D) Inverse correlation between serum transaminase levels (n = 8 mice/group) or hepatic triglyceride levels (n = 8 mice/group) and log(M2/M1) gene expression ratio with CD163 (M2) and iNOS (M1) mRNA expression (from data in A and B). Hepatocyte steatosis (E) and apoptosis (F) were evaluated by Oil red-O and cleaved-caspase-3 staining in EtOH-treated AML12 hepatocytes incubated with conditioned medium from control, M1, or M2 Raw264.7 cells. Assays are representative of three experiments (scale bar = 50 μm) *P < 0.05.

Figure 2.

Alcohol-fed BALB/c mice display preponderant M2 KC polarization and selective M1 KC apoptosis. (A) KC polarization was characterized in BALB/c and C57BL6/J chronically fed control diet (CD) or alcohol, by confocal microscopy (n = 12 mice/condition, two experiments); M1 (F4/80+/iNOS+) or M2 (F4/80+/CD206+) KCs are indicated by white arrows; F4/80+-CD206--iNOS- cells were classified as M0. Scale bar = 5 μm. (B) Apoptosis of KCs was characterized by confocal microscopy, using cleaved-caspase-3 and F4/80 antibodies (white arrows) in EtOH-fed BALB/c and C57BL6/J mice (n = 12/group, two experiments) (scale bar = 5 μm). (C,D) The phenotype of apoptotic KC (F4/80+-cleaved-caspase-3+) was characterized by triple immunohistofluorescence using iNOS (M1) or CD206 (M2) (scale bar = 2 μm). (E) Apoptosis of M1 (F4/80+/iNOS+) KCs was further characterized using TUNEL+ nuclei detection (scale bar = 2 μm) *P < 0.05.

We next compared the macrophage phenotype of the two strains. Alcohol-fed C57BL6/J mice showed a 3- to 9-fold induction in hepatic M1 genes (inducible nitric oxide synthase [iNOS], tumor necrosis factor alpha [TNFα], and MCP1), whereas M2 markers (Arginase 1 [Arg1], mannose receptor C type 2 [Mrc2], and cluster of differentiation 163 [CD163]) were unchanged or slightly increased (Fig. 1A). In contrast, BALB/c mice showed no change in the hepatic expression of M1 genes in response to alcohol, but displayed a higher hepatic expression of M2 markers, including Arg1, Mrc2, and CD163, both in control and alcohol feeding conditions (Fig. 1A).

M1-responsive C57BL6/J mice displayed significant steatosis, hepatocyte apoptosis, and elevation of serum transaminase levels, whereas M2 preponderant BALB/c mice were resistant (Fig. 1B,C). Analysis of pooled data from both strains of alcohol-fed mice further showed an inverse correlation between the ratio of M2/M1 mRNA expression and liver triglyceride levels or serum transaminase (Fig. 1D). In keeping with the in vivo data, conditioned medium from lipopolysaccharide (LPS)-stimulated M1-polarized macrophages promoted hepatocyte steatosis (Fig. 1E) and apoptosis (Fig. 1F), whereas interleukin (IL)4-stimulated (M2) conditioned medium had no effect. Altogether, these results indicate that alcohol-fed C57BL6/J mice display a predominant M1 response associated with steatosis and liver injury. In contrast, alcohol-fed BALB/c mice are characterized by preponderant M2 KC polarization, an impairment of the M1 response, and resistance to alcohol-induced liver injury.

Alcohol-Resistant BALB/c Mice Show Selective Hepatic M1 Macrophage Apoptosis

Macrophage phenotype was further characterized by double immunohistofluorescence, combining the macrophage marker F4/80 and either the M1 marker iNOS, or the M2 marker mannose receptor CD206. F4/80+ cells that expressed neither CD206 nor iNOS were classified as M0. Control C57BL6/J and BALB/c mice both exhibited a mixed hepatic population of M0/M1/M2 polarized macrophages (Fig. 2A). However, control BALB/c mice displayed a higher proportion of M2 macrophages, as compared to control C57BL6/J mice (40% versus 20% F4/80+/CD206+ cells, respectively, Fig. 2A). Intriguingly, chronic alcohol feeding of BALB/c mice caused a marked drop in the total number of KCs, as assessed by mRNA expression and F4/80 immunostaining (Figs. 1A, 2A), associated with a reduction in both M1 and M0 KC density (Fig. 2A). Residual KCs adopted a preponderant M2 polarization (60% of F4/80+/CD206+ cells in alcohol-exposed BALB/c mice (Fig. 2A). In contrast, alcohol did not modify the density of KCs in C57BL6/J mice, but promoted predominant M1 polarization (60% F4/80+/iNOS+ cells), a decrease in M0 KCs, with no change in the proportion of M2 KCs. Differential polarization adopted by alcohol-fed BALB/c and C57BL6/J KCs was confirmed by flow cytometry analysis (Fig. S2). F4/80high/CD206+ M2 cells represented 86% of total F4/80high cells in BALB/c mice but only 34% in C57BL6/J mice (Fig. S2).

Chronic alcohol feeding caused a 3-fold increase in KC apoptosis in BALB/c mice, as assessed by F4/80/cleaved-caspase-3 double immunostaining (Fig. 2B). Importantly, cleaved-caspase-3 staining was exclusively detected in F4/80+ cells (Fig. 2B), indicating that the apoptotic process selectively targets KCs in BALB/c mice, whereas there was no detectable caspase-3 signal in macrophages of alcohol-fed C57BL6/J mice (Fig. 2B). The phenotype of apoptotic KCs was further characterized by triple immunolabeling, combining F4/80, cleaved-caspase-3, iNOS, or CD206 antibodies. In alcohol-fed BALB/c mice, all cleaved-caspase-3+/ F4/80+ cells stained for iNOS, but remained CD206-, indicating selective M1 macrophage apoptosis (Fig. 2C,D). Similar results were obtained using terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) assay (Fig. 2E). Thus, alcohol-fed BALB/c mice are characterized by preponderant M2 KC polarization and M1 KC apoptosis.

M2 Macrophages Trigger Selective M1 Macrophage Apoptosis

The causal relationship between M2 KC polarization and the induction of M1 KC apoptosis was investigated in KCs isolated from C57BL6/J mice. M1 or M2 polarization was induced by incubation with LPS or IL4, respectively. Exposure of M1 KCs to conditioned medium from M2 KCs increased the number of cleaved-caspase-3-positive M1 KCs and decreased the density of M1 KCs (Fig. 3A). Of note, M2 conditioned medium exclusively promoted apoptosis of M1 KCs, and did not affect nonpolarized control KCs (Fig. 3A).

Figure 3.

M2 macrophages trigger selective M1 macrophage apoptosis. (A) Conditioned medium from M2 KCs (IL4 CM), but not from M1 KCs (LPS CM) or control KCs (cont CM) triggers apoptosis and decreases density of cultured LPS-stimulated KCs. Data are representative of two experiments (scale bar = 20 μm). (B) Resveratrol (30 μM) and adiponectin (3 μg/mL) promote M2 polarization; typical experiments reproduced twice. (C) Conditioned media from IL4-, adiponectin-, and resveratrol-treated M2 macrophages (Raw264.7 cells, left, KCs isolated from C57BL6/J mice, right) increase apoptosis and decrease density of LPS-stimulated macrophages. (D,E) Treatment of EtOH-fed C57BL6/J mice with resveratrol (400 mg/kg/day) (n = 6 for CD and n = 15 for EtOH in each group) limits steatosis (D), as characterized by Oil red-O staining (scale bar = 50 μM) and cleaved-caspase-3+ hepatocyte number (red arrows) (E, right). Resveratrol-treated mice display M2 KC polarization and KC apoptosis. (E, left): KC polarization was characterized in livers by triple F4/80, iNOS (M1), and CD206 (M2) immunofluorescence. (E, right): apoptotic macrophages (white arrows) were cleaved-caspase-3+/F4/80+ cells (scale bar = 20 μM). *P < 0.05 versus corresponding control. &P < 0.05 for C57BL6/J resveratrol EtOH versus C57BL6/J EtOH.

We then investigated whether other M2 inducers may trigger M2-induced apoptosis of M1 macrophages and focused on adiponectin and resveratrol, which have been shown to protect against alcohol-induced liver lesions[20-22] (Fig. 3D,E). We found that resveratrol and adiponectin up-regulate M2 gene expression in macrophages (Fig. 3B). Furthermore, the conditioned medium of macrophages exposed to adiponectin or resveratrol increased the proportion of caspase-3 positive M1 macrophages and decreased their survival (Fig. 3C). Noticeably, direct addition of either IL4, resveratrol, or adiponectin had no apoptotic effects (Fig. 3C), demonstrating that a soluble mediator released by M2 macrophages triggers selective apoptosis of M1 counterparts. In keeping with in vitro data, alcohol-fed C57BL6/J mice treated with resveratrol showed decreased M1 KC density and enhanced KC apoptosis, while the number of M2 KCs was increased (Fig. 3E; Table S1).

M2-Induced M1 Macrophage Apoptosis Involves Paracrine Activation of Arginase by Way of IL10

Recent studies have shown that activation of arginase may drive apoptosis of iNOS-expressing cells.[23] Addition of the arginase inhibitor NOR-NOHA to LPS-stimulated Raw264.7 macrophages prevented the appearance of caspase-3-positive signals elicited by IL4 (Fig. 4A) or resveratrol (Fig. 4B) conditioned media. In addition, NOR-NOHA limited the loss of cells with long spindle-shaped morphology, typically emerging in response to LPS-induced M1 polarization (Fig. S3A). Interestingly, apoptotic M1 Raw264.7 macrophages exposed to IL4-conditioned medium were characterized by a high coexpression of Arg1 and iNOS (Fig. 4C). In keeping with that, livers of alcohol fed BALB/c also showed high Arg1/iNOS coexpression that was exclusively detected in apoptotic KCs (Fig. 4C).

Figure 4.

M2 macrophages trigger selective M1 macrophage apoptosis by way of arginase activation. (A,B) Conditioned medium from M2 Raw264.7 macrophages (IL4 CM or resveratrol CM) induces apoptosis of LPS-stimulated macrophages (scale bar = 20 μm): cleaved-caspase-3+ cells (red arrows). When indicated, the arginase inhibitor NOR-NOHA (100 μm) was preincubated with LPS-stimulated macrophages, prior to addition of IL4 or resveratrol CM. Data are representative of three experiments. *P < 0.05 versus control LPS-stimulated macrophages. (C) iNOS+/Arg1+ phenotype of apoptotic macrophages of (left): LPS-stimulated Raw264.7 cells incubated with IL4 CM and (right): livers from EtOH-fed BALB/c mice (scale bar = 5 μm).

It has been reported that IL10 induces Arg1 expression in bone marrow-derived macrophages.[24] We determined whether this M2-secreted cytokine might mediate arginase-dependent apoptosis of M1 macrophages. Exposure of M1 cells to IL10 increased caspase-3-positive cell density and reduced spindle-shaped cell number (Fig. 5A; Fig. S3B). Moreover, the arginase inhibitor NOR-NOHA impaired IL10-induced cell death (Fig. 5A; Fig. S3B). Finally, anti-IL10 antibodies blunted apoptosis of LPS-stimulated M1 macrophages elicited by IL4 (Fig. 5B; Fig. S3C), resveratrol (Fig. 5C), and adiponectin (Fig. 5D) conditioned media. Experiments in LPS- or IL4-treated isolated peritoneal macrophages further confirmed that IL10 released by M2 macrophages triggers apoptosis of M1 cells by way of arginase activation (Fig. S4). In keeping with in vitro data, neutralization of IL10 by in vivo administration of an IL10 antibody prevented KC apoptosis and the related decrease in KC density during early alcohol feeding (Fig. 5E; Table S1), a situation that has recently been associated with enhanced macrophage apoptosis[25] and in which we found increased hepatic IL10 levels (Fig. 5F). Macrophage apoptosis was selective for M1 cells in these mice (Fig. 5E) and associated with predominant M2 polarization (Fig. S5).

Figure 5.

M2 macrophages trigger selective M1 macrophage apoptosis by way of IL10 secretion. (A-D) IL10, IL4, resveratrol, adiponectin CM induce apoptosis and decrease survival of LPS-stimulated macrophages (scale bar = 20 μm): cleaved-caspase-3+ cells (red arrows). When indicated, the arginase inhibitor NOR-NOHA (100 μm) (A) or a neutralizing anti-IL10 antibody (10 μg/mL) (B) were preincubated with LPS-stimulated Raw264.7 cells, prior to addition of IL10 (A) or of the conditioned medium from IL4 (B) or resveratrol (C) or adiponectin (D) treated cells. Data are representative of three experiments. *P < 0.05 versus control LPS-stimulated Raw264.7 cells. (E) C57BL6/J mice were injected with anti-IL10 or IgG1 isotype 1 hour prior to alcohol or isocaloric maltodextrin ingestion, and sacrificed 6 hours later (n = 8 mice/group). Apoptotic KCs were characterized by cleaved-caspase-3/F4/80 immunohistofluorescence. *P < 0.05 versus corresponding control. (E, right): iNOS+/Arg1+ phenotype of apoptotic macrophages of alcohol-exposed C57BL6/J mice. (F) IL10 levels were quantified in the livers of C57BL6/J mice 3 or 6 hours after alcohol ingestion (n = 12-30/group from two experiments). *P < 0.05 versus corresponding control.

High Hepatic M2 Marker Expression and KC Apoptosis Differentiate Heavy Alcohol Drinkers With Limited Liver Injury From Those With More Severe Lesions

We investigated the relationship of hepatic M2 signature with the severity of liver injury in liver biopsies obtained from heavy ongoing alcohol drinkers with limited fibrosis (Table 1). Patients were classified into two groups with minimally increased and elevated transaminases values at liver biopsy. There were no differences in age, daily alcohol intake, or duration of alcohol intoxication between groups (Table 1). Compared to patients with elevated transaminases (group 2), patients with minimally increased transaminases values (group 1) displayed limited hepatic injury, as shown by a lower grade of liver steatosis and a lower serum level of caspase-generated keratin 18 fragment, a biomarker of hepatocyte apoptosis in ALD[18] (Table 1). Determination of liver M2 and M1 signatures showed a higher mRNA expression of the M2 markers CD206 and CD163 in patients with limited hepatic injury (group 1), compared to those with more severe lesions (group 2) (Fig. 6A), whereas expression of IL10 and of the M1 marker TNF-α was similar in both groups (not shown). Cleaved-caspase-3 immunostaining disclosed negligible hepatocyte apoptosis in patients from group 1, compared to group 2 (Fig. 6B), and showed that positive hepatocytes were mostly circumscribed in steatotic foci (Fig. 6B). Surprisingly, cleaved-caspase-3 signal was also detected in nonparenchymal cells, distant from steatotic foci, that were identified as macrophages by CD68 costaining (Fig. 6B). Finally, cleaved-caspase-3 positive macrophages were more frequently detected in patients with minimally increased transaminases values and negligible hepatocyte apoptosis (group 1) (Fig. 6B).

Figure 6.

High hepatic expression of M2 markers is associated with macrophage apoptosis and limited lesions in biopsies of ongoing alcohol abusers. (A) M2 marker mRNA expression is higher in liver biopsies of heavy alcohol drinkers with limited (n = 10, group 1) versus more severe (n = 5, group 2) hepatic injury. (B) Left panel: typical images of (a) hepatocyte or (b) macrophage apoptosis in liver tissue sections stained with cleaved-caspase-3, or (c) of CD68+ macrophages. (d,e) Double cleaved-caspase-3/CD68 staining, with magnification of an apoptotic macrophage (inset d) and apoptotic hepatocyte (inset e) (scale bar = 50 μM). Right panel shows quantification of cleaved-caspase-3+ macrophages and hepatocytes. *P < 0.05.

High Hepatic M2 Marker Expression and KC Apoptosis Characterize Morbidly Obese Patients With Limited Hepatic Injury and High Fat-Fed Mice Cured With Resveratrol

We investigated the relevance of our findings in the context of NAFLD, both in mice and humans. C57BL6/J mice fed HFD for 27 weeks received curative treatment with resveratrol for the last 3 weeks (Table S1). HFD livers displayed increased F4/80+ cell density and preponderant M1 KC polarization (Fig. 7A). Administration of resveratrol to HFD-fed mice decreased KC density back to normal levels, switched KC polarization towards a preponderant M2 phenotype, and triggered M1 KC apoptosis (Fig. 7C). Moreover, administration of resveratrol strongly improved NAFLD as shown by the near complete inhibition of hepatocyte apoptosis (Fig. 7C) and steatosis (Fig. 7B).

Figure 7.

Curative treatment with resveratrol promotes M2 KC polarization, M1 KC apoptosis, and improves HFD-induced hepatocyte apoptosis and steatosis in mice. High hepatic expression of M2 markers is associated with macrophage apoptosis and limited lesions in biopsies of morbidly obese patients. C57BL6/J mice were fed normal diet (ND) or HFD for 27 weeks, with or without resveratrol (400 mg/kg/day) during the last 3 weeks (n = 6 and 10 mice per ND and HFD condition, respectively). (A) KC polarization was characterized by fluorescent microscopy. (B) Steatosis was evaluated by Oil red-O staining (scale bar = 50 μm). (C) Apoptosis of M1 macrophages and hepatocytes was characterized by triple iNOS, cleaved-caspase-3, and F4/80 immunofluorescence. *P < 0.05 versus ND; &P < 0.05 for C57BL6/J resveratrol HFD versus C57BL6/J HFD. (D) M2 marker mRNA expression is higher in liver biopsies of morbidly obese patients with limited (n = 7, group S0) versus more severe (n = 6, group S2) hepatic steatosis. (E) Hepatocyte or macrophage apoptosis was evaluated by double cleaved-caspase-3/CD68 staining in liver tissue sections from S0 and S2 patients (scale bar = 50 μM). Right panel shows quantification of cleaved-caspase-3+ macrophages and hepatocytes. *P < 0.05.

We also compared T-cell activation markers in HFD and alcohol-fed mice treated with resveratrol. Resveratrol increased hepatic Th1 gene expression in HFD-fed mice, whereas it decreased Th1 markers in alcohol-exposed treated animals; Th2 and Th17 gene expression were not modified (not shown). These results demonstrate that alcohol-fed and HFD-fed mice treated with resveratrol differ in Th1 environment, but display similar preponderant M2 signature and induction of M1 apoptosis. These data do not argue for a contribution of T cells in M1 apoptosis.

We examined the relationship of hepatic M2 signature to the severity of steatosis in liver biopsies obtained from morbidly obese patients undergoing bariatric surgery (Table 1). Patients were classified into two groups, with minimal (S0) and elevated (S2) steatosis. S0 patients showed a higher mRNA expression of the M2 markers CD206 and CD163 as compared to S2 patients (Fig. 7D), whereas the expression of IL10 and that of the M1 marker TNF-α was similar in both groups (not shown). Cleaved-caspase-3/CD68 positive macrophages were detected in liver biopsy of S0 and S2 patients but was more frequent in S0 patients, who showed negligible hepatocyte apoptosis (Fig. 7E).


Activation of Kupffer cells to secrete proinflammatory mediators is a key event in the initiation of fatty liver disease, and limiting their polarization into an M1 phenotype is considered an attractive strategy.[12, 26] In the present study, combining human data, animal models, and cell culture experiments, we identify a novel mechanism neutralizing M1 Kupffer cell emergence, which relies on selective induction of their apoptosis by M2 Kupffer cells.

The successful resolution of inflammatory processes requires the inhibition of proinflammatory signaling. M2 macrophages typically fulfill this function, owing to their high capacity to counteract the proinflammatory functions of classical macrophages (M1).[1, 2] We postulated that favoring M2 KC polarization might protect against fatty liver disease. The relevance of this hypothesis was evaluated in liver biopsies from either ongoing alcohol abusers or morbidly obese patients, with mild forms of ALD or NAFLD, and classified according to the degree of liver lesions. Individuals with limited liver lesions displayed higher hepatic M2 gene expression and negligible hepatocyte apoptosis, as compared to patients with more severe lesions. These data provided a link between M2 KC polarization and the prevention of fatty liver disease against progression to more severe forms of injury. Moreover, they raise the intriguing possibility that differences in Kupffer cell phenotype might account for the variability in susceptibility of individuals to ALD or NAFLD, in addition to incriminated environmental, genetic, and metabolic factors.[27, 28] We also investigated the relationship between M2 KC polarization, prevention, or regression of fatty liver injury in mice models. Genetic or pharmacological interventions favoring preponderant M2 KC polarization (i.e., BALB/c mice fed alcohol, and resveratrol-treated C57BL6/J mice fed either alcohol or high fat) were associated with impaired M1 response and limited liver injury. The ability of M2 Kupffer cells to produce fibrogenic mediators has raised concerns as to a potential deleterious impact on liver fibrogenesis. However, our results show that promotion of preponderant M2 KC polarization in alcohol or high fat fed mice do not enhance fibrogenic gene expression (Fig. S6). Although additional investigations are needed to clarify the role of the M1/M2 Kupffer cell balance in the control of liver fibrosis, it should be noted that several recent studies have documented antifibrogenic properties of M2 macrophages.[29]

Interestingly, in alcohol-fed BALB/c mice the emergence of M2 KC occurred in the absence of recruitment of Gr-1 expressing monocytes, and without evidence for KC proliferation, as assessed by bromodeoxyuridine (BrdU) staining (Fig. S7). These results challenged the assumption that accumulation of M2 macrophages results from the recruitment of circulating monocytes at sites of injury[1, 2] or arises from resident macrophages undergoing in situ proliferation.[30] Our data rather suggest that the emergence of M2 KC in alcohol-fed BALB/c mice may occur at the expense of nonpolarized resident M0 macrophages that markedly decrease in number upon chronic alcohol feeding.

Identification of M1 KC apoptosis by their M2 counterparts constitutes a major point of our study. Kupffer cell apoptosis has been recently described as a feature of early alcohol response.[25, 31] Interestingly, we detected macrophage apoptosis in the liver of heavy alcohol drinkers or morbidly obese patients, and observed that macrophage death was preponderant in individuals with mild liver injury and predominant M2 signature. Animal studies also highlighted that alcohol- or high fat-fed mice with preponderant M2 KC polarization displayed enhanced KC apoptosis, and limited liver injury. The apoptotic response was restricted to M1-polarized KC and was not detected in other hepatic cell types. These data revealed a positive relationship between M2 KC polarization and M1 macrophage apoptosis, and led us to postulate that M2 KCs might induce M1 macrophage apoptosis. Conditioned medium experiments demonstrated that several pro-M2 stimuli induce M1 macrophage apoptosis. Indeed, macrophages polarized into an M2 phenotype by either IL4, adiponectin, or resveratrol displayed apoptotic properties selectively targeting M1 macrophages, without affecting resting M0 cells. Taken together, these data identify a new mechanism for M1 macrophage elimination that relies on M2-induced M1 macrophage apoptosis. They reveal an as yet unsuspected fratricide mechanism regulating the balance between M1 and M2 macrophages.

Mechanistically, we identify IL10 as the mediator of M1 Kupffer cell apoptosis induced by M2 counterparts. As described in macrophages from diverse origins, IL10 is secreted by M2 macrophages and displays potent anti-inflammatory properties,[1, 2, 21, 32] in particular in the context of ALD. Thus, IL10-deficient mice show enhanced sensitivity to alcohol-induced liver injury.[32] Moreover, IL10 suppresses LPS-stimulated TNFα expression in KC after chronic alcohol feeding.[21] We found that IL10 promotes M1 apoptosis, whereas anti-IL10 antibodies blunted the proapoptotic effects of IL4, resveratrol, and adiponectin conditioned media. In keeping with in vitro data, administration of anti-IL10 antibodies reduced KC apoptosis in alcohol-fed mice. While the secretion of IL10 is largely associated with the dampening of many apoptotic stimuli,[33] our results highlight a proapoptotic action of IL10 targeting M1 macrophages. We also demonstrate that the mechanism of IL10-induced M1 macrophage death relies on activation of arginase activity in high iNOS expressing cells, as shown in vitro by triple immunolabeling, and by the blockade of macrophage apoptosis upon pharmacological inhibition of arginase activity. These results identify arginase as a novel apoptotic pathway for IL10 in macrophages. A possible mechanism underlying these apoptotic effects may be that, following combined activation of arginase and iNOS, the resulting competition for their common substrate L-arginine leads to decreased arginine availability and to a switch of iNOS function towards proapoptotic properties. Indeed, low levels of arginine down-regulate NO synthesis but enhance O-2 production by iNOS, generating proapoptotic peroxinitrites.[23] Whether arginase-dependent elimination of M1 by M2 KCs may contribute to the IL10-anti-inflammatory effects reported in various models of acute and chronic liver injury[32, 34-36] remains to be investigated.

In summary, our results reveal an as yet unrecognized mechanism limiting M1 KC functions that depends on proapoptotic effects of M2 KCs towards their M1 counterparts, by way of IL10-dependent paracrine interactions. They suggest that pharmacological interventions targeting M2 KC polarization during the early stages of ALD and NAFLD may represent an attractive strategy for the limitation of inflammation and hepatocyte injury.


We thank Anne Hulin and Irina Andriamanana, from the Toxicology Department of the Henri Mondor Hospital, for serum ethanol measurement, Adeline Henry and Aurélie Guguin, from the cytometry platform, for flow cytometry analyses, Xavier Ducroy from the Imaging platform for confocal image capture, and Sophia Balustre for help during in vivo experiments.

Author Contributions

JW, FT-C, AL, FP, AT, PG, AM, SL, and CP: study concept and design; JW, MB, FT-C, AL, SB, FL, FP, AT, PG, and CP: acquisition of data; JW, FT-C, AL, FP, AT, PG, SL, and CP: analysis and interpretation of data; FP, PG, AM, SL, and CP: drafting and critical revision of the article for important intellectual content; CP: statistical analysis; SL obtained funding; SL and CP: study supervision.