Hepatic macrophages but not dendritic cells contribute to liver fibrosis by promoting the survival of activated hepatic stellate cells in mice


  • Potential conflict of interest: Nothing to report.

  • This study was supported by NIH grants R01DK076920 and U54CA163111 (both to Robert F. Schwabe). Jean-Philippe Pradere was supported by a postdoctoral fellowship from the American Liver Foundation. Johannes Kluwe was supported by the German Research Foundation (grant KL2140/2-1) and a Sheila Sherlock fellowship from the European Association for the Study of the Liver. Ingmar Mederacke was supported by the German Research Foundation (grant ME3723/1-1). Dianne Dapito was supported by NIH grant 1F31DK091980. Costica Aloman was supported by NIH grant 1K08DK088954.

Address reprint requests to: Robert F. Schwabe, Department of Medicine, Columbia University, College of Physicians & Surgeons, Russ Berrie Pavilion, Room 415, 1150 St. Nicholas Ave, New York, NY 10032; E-mail: rfs2102@columbia.edu; fax: 212-851-5461


Although it is well established that hepatic macrophages play a crucial role in the development of liver fibrosis, the underlying mechanisms remain largely elusive. Moreover, it is not known whether other mononuclear phagocytes such as dendritic cells (DCs) contribute to hepatic stellate cell (HSC) activation and liver fibrosis. We show for the first time that hepatic macrophages enhance myofibroblast survival in a nuclear factor kappa B (NF-κB)–dependent manner and thereby promote liver fibrosis. Microarray and pathway analysis revealed no induction of HSC activation pathways by hepatic macrophages but a profound activation of the NF-κB pathway in HSCs. Conversely, depletion of mononuclear phagocytes during fibrogenesis in vivo resulted in suppressed NF-κB activation in HSCs. Macrophage-induced activation of NF-κB in HSCs in vitro and in vivo was mediated by interleukin (IL)−1 and tumor necrosis factor (TNF). Notably, IL-1 and TNF did not promote HSC activation but promoted survival of activated HSCs in vitro and in vivo and thereby increased liver fibrosis, as demonstrated by neutralization in coculture experiments and genetic ablation of IL-1 and TNF receptor in vivo. Coculture and in vivo ablation experiments revealed only a minor contribution to NF-κB activation in HSCs by DCs, and no contribution of DCs to liver fibrosis development, respectively. Conclusion: Promotion of NF-κB–dependent myofibroblast survival by macrophages but not DCs provides a novel link between inflammation and fibrosis. (Hepatology 2013;58:1461–1473)


α-smooth muscle actin


bile duct ligation


carbon tetrachloride


dendritic cells


classical dendritic cell


double knockout


fluorescence-activated cell sorting


green fluorescent protein


hepatic macrophages


hepatic stellate cell




Ingenuity Pathway Analysis


messenger RNA


nuclear factor kappa B


phosphate-buffered saline


plasmacytoid DC


quantitative real-time polymerase chain reaction


tumor necrosis factor


terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling.

The development of liver fibrosis constitutes one of the major complications of chronic liver disease, with many clinical consequences such as the development of esophageal varices and ascites being directly related to the presence of liver fibrosis.[1] The hepatic wound healing response is a concerted action of multiple resident and nonresident cell types that not only provides a scaffold for structural stability but also involves the removal of cellular debris by infiltrating hepatic macrophages (HM) and the regeneration of functional parenchyma.[2, 3] Hepatic stellate cells (HSCs) are considered the main fibrogenic cell type in the liver and are responsible for the production of various types of extracellular matrix.[2, 3] HSCs undergo a well-characterized activation process, during which they lose their characteristic vitamin A and lipid stores and obtain a myofibroblastic phenotype.[2, 3] The activation of HSCs is controlled by multiple soluble mediators, including transforming growth factor β and platelet-derived growth factor, and is part of a complex cellular network that controls the hepatic wound healing response. Previous studies have demonstrated that multiple cell populations—including HMs, myeloid-derived suppressor cell, B cells, T cells, and natural killer cells—influence the development of liver fibrosis.[4-12] Among those, HMs exert a profound effect on HSCs and hepatic fibrosis as shown by genetic or pharmacologic models of macrophage depletion.[6, 7, 13] At the same time, HMs also contribute to fibrosis resolution through MMP13 and matrix remodeling.[6, 14, 15] However, the mechanisms by which HMs promote liver fibrosis remain largely elusive. Dendritic cells (DCs) are developmentally closely related to macrophages and exert a profound effect on liver fibrosis regression[16] and the cytokine microenvironment during fibrogenesis,[12] but their contribution to liver fibrosis development remains unknown.

In the present study, we uncovered the promotion of HSC/myofibroblast survival as a novel mechanism through which macrophages promote fibrosis. Moreover, we demonstrate for the first time that neither classical DCs (cDCs) nor plasmacytoid DCs (pDCs) contribute to fibrogenesis.

Materials and Methods

Mice and Liver Fibrosis Induction

C57BL/6 mice, Balb/c mice, CD11c-DTR-eGFP mice (in C57Bl/6 background), Tnfrsf1a/Il1r1-deleted (“dko”) mice (in B6.129S background), and B6.129S mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and housed in a specific pathogen-free facility. Collagen–green fluorescent protein (GFP) reporter mice have been described.[17] Hepatic fibrosis was induced in 8- to 12-week-old male mice by ligating the common bile duct for 5 to 15 days as described,[18, 19] via 4 to 20 intraperitoneal injections of carbon tetrachloride (CCl4) (0.125 μL/g to 0.5 μL/g body weight, all dissolved in corn oil at a ratio of 1:3) or 18-week treatment with 300 mg/L thioacetamide in drinking water as described.[13] Some mice received single or repeated intraperitoneal injections of 200 μL liposomal clodronate (5 mg/mL) or liposomal vehicle as described.[13] All animal procedures were approved by the Columbia University or Mount Sinai School of Medicine Institutional Animal Care and Use Committee, and were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

DC Depletion

All cDC depletion studies were performed in CD11c-DTR chimeric mice expressing CD11c-DTR only in bone marrow and its progeny. In the bile duct ligation (BDL) fibrosis model, cDC depletion was achieved via two intraperitoneal injections of diphtheria toxin (25 ng/g body weight) or phosphate-buffered saline (PBS) at days 4 and 6. In the CCl4 fibrosis model, depletion of cDC was achieved via intraperitoneal diphtheria toxin injection every 72 hours, 25 ng/g for the first 2 weeks followed by 10 ng/g for the last 2 weeks. For the depletion of pDC, C57B6 mice were injected with pDC-depleting antibody 120G8 or isotype control (500 μg/mouse IP dissolved in 200 μL saline) every 48 hours during the last 2 weeks of CCl4-induced fibrosis.

Statistical Analysis

All data are expressed as the mean ± SD. For comparison of two groups, a two-sided unpaired t test or Mann-Whitney test were used. For multiple group comparisons, analysis of variance with Tukey post hoc analysis was performed. For correlation, the Pearson correlation coefficient was calculated. P < 0.05 was considered statistically significant.

Additional procedures are described in the Supporting Information.


Macrophages Activate Nuclear Factor kappa B in HSCs In Vitro and In Vivo

HSCs activate in a complex in vivo environment, characterized by the presence of multiple resident and recruited cell populations, including macrophages. To identify signaling pathways through which HMs exert profibrogenic effects, we determined via microarray analysis which genes and signaling pathways are activated in HSCs cocultured with F4/80-positive HMs from fibrotic livers (Supporting Fig. 1). Microarray analyses revealed that coculture of HSCs with HMs in a contact-independent manner resulted in a profound influence on gene expression, shifting the pattern toward those observed in in vivo–activated HSCs isolated either from bile duct–ligated or CCl4-treated mice (Fig. 1A,B), as previously postulated by us.[18] Ingenuity Pathway Analysis (IPA) of the more than 1,400 genes with significant and >2-fold change (Supporting Table 1) revealed liver fibrosis and inflammatory responses to be the most significant toxicological and biological functions (Supporting Fig. 2A,B), and the nuclear factor kappa B (NF-κB) pathway to be the center component of the highest-ranked network (Fig. 1C). Accordingly, NF-κB–regulated genes were significantly overrepresented among genes with more than 10-fold induction (chi-squared test; P < 0.00001). Out of the 82 genes with more than 10-fold up-regulation, 18 have been described as NF-κB–dependent (Fig. 1D and Supporting Table 2). We confirmed these results via real-time qRT-PCR and found that coculture with HMs for 24 hours or for 5 days increased the expression of known NF-κB–regulated genes, including Il6, Saa3, Cxcl5, Cxcl14, Serpinb2, Ch25h, and Mmp13 (Fig. 2A,C). Surprisingly, HMs did not induce classical HSC activation markers such as Col1a1, Col1a2, or Acta2 messenger RNA (mRNA) and did not change α-smooth muscle actin (α-SMA) protein levels in HSCs (Fig. 2C). We confirmed that all NF-κB–dependent genes, including Timp1, were suppressed in the presence of adenoviral IκB superrepressor (Fig. 2A) or by short-term treatment with IKK inhibitor Bay 11-7085 at very low nontoxic concentrations (Supporting Fig. 3). NF-κB activation was further confirmed via p65 immunohistochemistry (Fig. 2D) and immunoblot (Fig. 2E) demonstrating p65 translocation, p65-S536 phosphorylation, and IκBα degradation in HSCs treated with conditioned media from HMs but not after treatment with control media. Similar observations were made in an NF-κB reporter assay, in which coculture with HMs induced a >15-fold increase in NF-κB–driven luciferase activity (Fig. 2F). Based on these results, we focused on the NF-κB pathway in subsequent analyses of mechanisms by which HMs affect HSCs and fibrogenesis. Next, we determined whether HMs alter NF-κB–dependent gene expression in HSCs in the fibrotic liver by employing a depletion approach. For this purpose, we analyzed gene expression in fluorescence-activated cell sorting (FACS) ultrapure HSCs isolates that were immediately lysed after isolation and thus provide a “snapshot” of HSC gene expression in the fibrotic liver. NF-κB–dependent gene expression was highly up-regulated in HSCs activated in vivo compared with quiescent HSCs (Fig. 2G). Macrophage depletion by repeated liposomal clodronate injection efficiently reduced F4/80-positive and CD11b- and F4/80-double positive macrophages and ameliorated liver fibrosis following BDL and CCl4 treatment (Supporting Fig. 4). Notably, macrophage depletion strongly suppressed the expression of the NF-κB–dependent genes that were up-regulated by HMs in our coculture system (Fig. 2G). We further excluded that liposomal clodronate directly affects NF-κB via NF-κB reporter assay and or cell death in cultured HSCs (Fig. 2H,I).

Figure 1.

Microarray and pathway analysis reveal NF-κB and not fibrogenic activation as the predominant effect of hepatic macrophages on HSCs. (A, B) Heatmaps from microarrays showing quiescent HSCs, 5-day in vitro–activated HSCs, 5-day in vitro–activated HSCs cocultured with HMs, HSCs from mice bile duct–ligated for 15 days, and HSCs from mice after 4 CCL4 injections include (A) all genes that were significantly (adjusted P < 0.05) and at least four-fold up- or down-regulated in HSCs from BDL livers and (B) all genes that were significantly (adjusted P < 0.05) and at least four-fold up- or down-regulated in HSCs from CCl4 livers. (C) The highest-ranking network from IPA analysis. (D) Heatmap of NF-κB–responsive genes with significant (adjusted P < 0.05) and at least 10-fold induction in response to coculture with HMs.

Figure 2.

Hepatic macrophages induce NF-κB but not myofibroblastic activation in HSCs in vitro and in vivo. (A, B) HSCs were cocultured with HMs for 5 days (A), followed by determination of NF-κB–responsive genes via qRT-PCR or western blot analysis for α-SMA (B). (C) HSCs were infected with AdIκBsr or an empty control adenovirus followed by coculture with HSCs for 24 hours. mRNA expression of HSC activation markers and NF-κB–responsive genes was determined via qRT-PCR. (D) HSCs were stimulated with conditioned media from HMs for 1 hour. NF-κB activation was determined via immunofluorescent p65 staining (red) in HSCs costained for F-actin (green). Nuclei were marked with Hoechst stain (blue). (E) HSCs were stimulated with conditioned media from HMs for various times. NF-κB activation was determined via immunoblotting for S536-phosphorylated p65 and IκBα. (F) HSCs were transduced with an adenoviral NF-κB reporter. After coculture with HMs for 24 hours, luciferase activity was determined and is expressed as fold induction compared with HSCs cultured with empty inserts. (G) Gene expression patterns were determined in ultrapure unplated HSCs isolated from 15-day bile duct–ligated vehicle-treated mice (n = 6 isolations) and clodronate-treated mice (n = 4 isolations) mice via a combination of gradient centrifugation and FACS without additional plating. mRNA expression of fibrogenic genes and NF-κB–dependent genes was determined via qRT-PCR and expressed in comparison to quiescent HSCs isolated from control wild-type mice (n = 3 isolations). (H, I) HSCs were treated with 50 μL of 5 mg/mL solution per mL of media or liposomal vehicle for 12 hours. TNF-α– and IL-1β–induced NF-κB reporter activity (H) and cell death, showing Hoechst-stained nuclei (blue), retinoids (green), and propidium iodide–positive nuclei (red) were determined (I). **P < 0.01 versus HSCs cultured without HMs. *P < 0.05 versus HSCs cultured in the absence of HM ##P < 0.01 versus AdShuttle-infected HSCs cocultured with HM. xP < 0.05, xxP < 0.01 versus HSC BDL Veh. ns, not significant.

Interleukin-1 and TNF Mediate Macrophage-Induced NF-κB Activation in HSCs

Next, we investigated mechanisms through which HMs induce NF-κB activation in HSCs. First, we tested the contribution of interleukin (IL)−1 and TNF to HM-induced NF-κB in HSCs based on their known potent activation of NF-κB, the presence of the IL-1 receptor in the NF-κB network identified by IPA analysis (Fig. 1C), and up-regulated M1 markers inducible nitric oxide synthase and Cox2 in HMs from BDL mice (Supporting Fig. 1C). HMs induced NF-κB to the same degree as rmIL-1β, and to a higher degree than rmTNFα (Fig. 3A). TNF and IL-1 neutralization in cocultures by antagonistic TNF-receptor I-Fc and IL-1 receptor I-Fc chimera demonstrated potent inhibition of NF-κB–driven luciferase activity by IL-1 neutralization and moderate inhibition by TNF neutralization (Fig. 3B). Combined TNF and IL-1 neutralization almost completely blocked the HM-induced NF-κB–driven luciferase activity. Similarly, IL-1 and TNF neutralization also blocked HM-induced up-regulation of the NF-κB–dependent genes Ch25h, Cxcl5, Saa3, Serpinb2, Il6, and Mmp13, with IL-1 neutralization exerting pronounced effects and TNF neutralization exerting moderate effects (Fig. 3C). Again, combined TNF and IL-1 neutralization resulted in almost complete suppression of NF-κB–dependent gene expression. Conversely, IL-1β and TNF-α up-regulated all NF-κB target genes in HSCs that were induced by HMs, with the exception of Cxcl14 (Fig. 3F). Moreover, converting the HM population that consisted of a mixed M1/M2 phenotype (Supporting Fig. 1C) as in previous studies,[20] to an inflammatory M1 phenotype by treatment with lipopolysaccharide and interferon-γ further increased the expression of NF-κB–dependent genes in HSCs (Fig. 3D). Conversely, converting the HM population to an M2 phenotype by combined treatment with IL-10 and IL-4 suppressed the expression of NF-κB–dependent genes in HSCs (Fig. 3E).

Figure 3.

Macrophage-derived IL-1 and TNF promote NF-κB activation but not myfibroblastic activation in HSCs. (A) NF-κB–driven luciferase activity was determined in HSCs that were cocultured with HMs or treated with IL-1β (5 ng/mL) or TNFα (30 ng/mL) for 24 hours. (B) NF-κB–driven luciferase activity was determined in HSCs cocultured with HMs in the presence of vehicle (0.1% bovine serum albumin in PBS), antagonistic TNFRI-Fc chimera (TFc, 0.5 μg/mL), antagonistic IL-1RI-Fc chimera (IL-Fc, 0.5 μg/mL), or both for 24 hours. (C) mRNA expression of HSC activation markers or NF-κB–responsive genes was determined via qRT-PCR in HSCs cocultured with HMs in the presence or absence of TNFRI-Fc chimera, IL-1RI-Fc chimera, or both for 24 hours. (D, E) Macrophages isolated from BDL livers were treated for 16 hours with lipopolysaccharide (10 ng/mL) and interferon-γ (10 ng/mL) to convert them into M1 phenotype (D), or with IL-4 (10 ng/mL) and IL-10 (10 ng/mL) to convert them into M2 phenotype (E). Following 5 washes, M1 or M2 macrophages were cocultured with HSCs. Coculture-induced expression of NF-κB–dependent genes was determined via qRT-PCR. (F) HSCs were treated with rmIL-1β (5 ng/mL) or rmTNFα (30 ng/mL) for 6 hours followed by analysis of HSC activation markers and NF-κB–dependent genes via PCR. (G) α-SMA levels were analyzed via immunoblotting after treating HSCs with rmIL-1β or rmTNF for 5 days. *P < 0.05. **P < 0.01. ns, not significant.

IL-1 and TNF Do Not Induce HSC Activation but Enhance HSC Survival In Vitro and In Vivo

Because one main function of the NF-κB pathway is protection from cell death, we next determined whether HM-induced NF-κB activation prevented HSC death or whether it directly promoted HSC activation. In contrast to previous studies,[21] IL-1β and TNF-α failed to directly induce HSC activation (Fig. 3F,G). Coculture of HSCs with HM strongly suppressed cell death induced by prolonged cell culture in low-serum media (Fig. 4A), a well-established method of inducing HSC death[22, 23] that showed signs of apoptosis such as caspase-3 cleavage (Fig. 4A). The protective effects of HM were almost completely abolished by the combined neutralization of IL-1 and TNF (Fig. 4A). Conversely, rmIL-1β was as efficient as HMs in rescuing HSCs from cell death (Supporting Fig. 5). Furthermore, neutralization of IL-1 and TNF did not reduce viability of HMs, and HM supernatant could also suppress HSC death (data not shown), again emphasizing that HSCs, not HMs, are the relevant targets of IL-1 and TNF. To determine whether this survival pathway was also responsible for the decreased fibrosis observed in macrophage-depleted mice, we performed terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling (TUNEL) assays in clodronate- or vehicle-treated collagen-GFP reporter mice following BDL to detect apoptosis in GFP-positive collagen-producing myofibroblasts. In mice receiving liposomal clodronate during fibrogenesis, we detected a five-fold increase in GFP- and TUNEL-double positive cells using confocal microscopy (Fig. 4B). Notably, the rate of TUNEL-positive HSCs in vehicle-treated mice were very similar to peak rates in previous studies that investigated the contribution of HSC apoptosis to fibrosis resolution,[22, 24] thus underscoring the relevance of this dramatic increase in HSC apoptosis for liver fibrosis. We found no effect of liposomal clodronate on HSC viability, thereby excluding the possibility that liposomal clodronate directly induces HSC apoptosis in fibrotic livers (Fig. 2I). We also observed reduced IL-1β and TNF-α mRNA in fibrotic livers from clodronate-treated mice (Fig. 4C). To test the in vivo relevance of this pathway, we first investigated how deficiency of IL-1β, the predominant activator of NF-κB in our coculture experiments, affects liver fibrosis. In contrast to previously published studies, we found no statistically significant difference in BDL-induced liver fibrosis between IL-1R1 knockout and wild-type mice, and further confirmed this data in the CCl4 and thioacetamide models of liver fibrosis (Supporting Fig. 6). If IL-1 signaling promoted liver fibrosis by increasing NF-κB–dependent HSC survival rather than direct HSC activation, it would be likely that TNF-α, the other major NF-κB–activating cytokine produced by macrophages, could still achieve NF-κB activation in HSCs and thus compensate for the loss of IL-1 signaling in this model. Based on the hypothesis that absence of both IL-1 and TNF signaling would be required to reduce HSC survival and liver fibrosis, we performed BDL in TNFR1/IL1R1 double knockout mice (dko) and wild-type control mice. Compared with wild-type mice, dko mice showed significantly reduced hepatic fibrosis after 5 or 15 days of BDL (Fig. 5A-B) and a five-fold increase in apoptotic TUNEL and desmin double-positive HSCs without significant differences in hepatic injury (Fig. 5B), supporting our hypothesis that suppression of both IL-1 and TNF signaling are required to affect HSC survival and liver fibrosis. Moreover, we found a significant reduction of NF-κB–dependent genes—including Il6, Cxcl5, Saa3, Serpinb2, and Timp1—in ultrapure unplated HSCs from dko mice, thus confirming that NF-κB activation in HSCs was mediated by TNF and IL-1 (Fig. 5C). Our microarray analysis revealed an up-regulation of two Trail decoy receptors, murine Trail decoy receptor 1 (Tntrsf23) and murine Trail decoy receptor 2 (Tnfrs22), in HSCs cocultured with HMs and in HSCs from BDL and CCl4 livers (Fig. 5D and Supporting Table 2). Notably, Trail-mediated apoptosis is a major contributor to HSC cell death induced by hepatic natural killer cells in vitro and in vivo.[11, 25] Neutralization of TNF or IL-1 prevented the up-regulation of Tnfrsf22 and Tnfrsf23 mRNA by HMs in coculture experiments (Supporting Fig. 7A). Moreover, depletion of HMs by liposomal clodronate or dko of TNFR1 and IL1R1 reduced Tnfrsf22 and Tnfrsf23 expression in vivo (Supporting Fig. 7B).

Figure 4.

HMs protect HSCs from cell death. (A) HSCs were starved in 0.1% fetal bovine serum in the presence or absence of HMs and the absence or presence of antagonistic TNFRI-Fc and IL-1RI-Fc chimera. Cell death was determined via propidium iodide staining (red). Nuclei were visualized via Hoechst staining (blue). Vitamin A–containing HSC lipid droplets were visualized via autofluorescence (green). Representative images and a quantification of the average of three independent HSC and HM isolations are shown. The inset shows cleaved caspase-3 immunofluorescent staining (red) and Hoechst (blue). (B) To determine apoptosis in collagen-expressing myofibroblasts, collagen-GFP reporter mice were treated with vehicle (n = 5) or clodronate (n = 4) followed by BDL for 9 days and TUNEL staining of liver sections. Cell death of collagen-expressing myofibroblasts was visualized via confocal microscopy and quantified by counting cells double-positive cells for GFP (green) and TUNEL (red). Nuclei were visualized by Hoechst staining (blue). (C) Hepatic levels of IL-1β and TNFα mRNA were determined via qRT-PCR in bile duct–ligated vehicle (n = 17) and clodronate-treated (n = 10) mice. *P < 0.05; **P < 0.01.

Figure 5.

TNF and IL-1 mediate NF-κB activation and protection from cell death in HSCs during liver fibrosis. (A) TNFRI and IL1-RI dko (n = 5) and wild-type mice (n = 10) underwent BDL for 14 days. Hepatic fibrosis was assessed via sirius red staining and polarized microscopy, α-SMA immunoblotting, and hydroxyproline assay. (B) TNFRI and IL1-RI dko (n = 7) and wild-type mice (n = 9) underwent BDL for 5 days. Hepatic fibrogenesis was assessed via qRT-PCR for profibrogenic genes, hepatic injury by ALT measurement. (C) Following 2 weeks of BDL, HSCs were isolated from mice TNFRI and IL1-RI dko (n = 5 isolations) or wild-type mice (n = 5 isolations) via gradient centrifugation and vitamin A fluorescence–based FACS without plating. mRNA expression of fibrogenic and NF-κB–dependent genes was determined via qRT-PCR and expressed in comparison with quiescent HSCs isolated from wild-type mice (n = 3 isolations). (D) To determine apoptosis in HSCs, liver sections from 14-day bile duct–ligated TNFRI and IL1-RI dko (n = 5) and wild-type mice (n = 5) were stained for desmin (green) and via TUNEL (red). Nuclei were marked via Hoechst staining (blue fluorescence). Cells positive for both desmin and TUNEL were quantified. (E) Heatmap of TNF receptor family members including Trail and Trail decoy receptors under different conditions, determined by microarray analysis (left panel). Expression of Trail decoy receptors Tnfrsf22 and Tnfrsf23 as determined via qRT-PCR in ultrapure unplated HSCs isolated from 14-day bile duct–ligated wild-type mice (n = 5 isolations) and double-deficient for TNFRI and IL1-RI mice (“Dko”, n = 5 isolations) via a combination of gradient centrifugation and FACS analysis without additional plating (center and right panel). Data are expressed as fold induction compared with quiescent HSCs isolated from control wild-type mice (n = 3 isolations) or as the ratio between the investigated mRNA and 18s mRNA if mRNA expression was not detectable in quiescent HSC. *P < 0.05; **P < 0.01. nd, not detectable.

DCs Do Not Contribute to HSC Activation and Fibrosis In Vivo

Liposomal clodronate does not affect HSC number[13] or biology (Fig. 2H,I), but may deplete DCs, a highly endocytotic cell population, as demonstrated by our FACS analysis (Fig. 6A). DCs contribute to liver inflammation[12] and to the regression of liver fibrosis,[16] but their role in the development of liver fibrosis is not known. To test whether DCs may contribute to HSC activation and liver fibrogenesis, we first performed a coculture of DCs and HSCs. Similar to HMs, DCs did not activate HSCs but rather up-regulated the expression of NF-κB–dependent genes, and NF-κB–driven luciferase reporter activity through an IL-1– and TNF-dependent manner (Fig. 6B). However, activation of NF-κB was considerably lower than the induction we observed in HM coculture. Based on these results, we next determined whether DC ablation may have contributed to the reduced fibrogenesis in clodronate-treated mice. In our first approach, we performed BDL in diphtheria toxin-treated or PBS-treated bone marrow–chimeric CD11c-DTR-eGFP mice. Bone marrow chimerism avoids the known side effects of diphtheria toxin treatment observed after long-term treatment in global CD11c-DTR-eGFP mice.[26] We did not observe a significant difference in BDL-induced fibrosis as determined by sirius red staining and qRT-PCR for the fibrogenic genes α-SMA, Col1a1, and TIMP1 (Fig. 6C-D). We confirmed these data employing CCl4 injection for induction of liver fibrosis, again using bone marrow-chimeric CD11c-DTR-eGFP mice. Similar to the BDL model, we did not observe significant differences in liver fibrosis between PBS and diphtheria toxin-treated mice (Fig. 6E). As a third approach, we used antibody-mediated ablation of pDC. Again, we did not observe a reduction of CCl4-induced liver fibrosis (Fig. 6F). Importantly, we achieved considerable depletion of cDC and pDC using the above methods (Supporting Fig. 8). Similar to previous studies,[27] we observed neutrophilia in CD11c-DTR mice (Supporting Fig. 9) but consider this unlikely to exert a profound effect on fibrosis based on previous studies.[28] Thus, our data suggest that neither class of DC significantly contributes to liver fibrogenesis in vivo.

Figure 6.

DCs moderately induce NF-κB–dependent gene transcription in HSCs but do not contribute to BDL- and CCl4-induced liver fibrosis. (A) Bile duct–ligated mice were injected every 5 days for a total of three injections with PBS (n = 4) or clodronate (n = 3) followed by quantification of cDCs using CD11c, and MHCII flow cytometry FACS plots show percentage of double-positive cells. (B) mRNA expression of HSC activation markers (left panel), NF-κB-responsive genes (middle panel), or NF-κB–driven luciferase activity (right panel) were determined via qRT-PCR and NF-κB reporter assay, respectively, in HSCs cocultured with DCs in the presence of antagonistic TNFRI-Fc and IL-1RI-Fc chimera (both 0.5 μg/mL) or vehicle (0.1% bovine serum albumin in PBS) for 24 hours. (C, D) Liver fibrosis was induced via BDL in CD11c-DTR bone marrow–chimeric mice. Chimeric male CD11c-DTR were treated with two injections of PBS (n = 5) or diphtheria toxin (25 ng/g body weight) (n = 6). Mice were sacrificed after 7 days. Deposition of fibrillar collagen was determined via sirius red staining (left upper panel) and morphometric quantification (right upper panel) (C). Expression of profibrogenic genes Acta2, Col1a1, and TIMP1 was determined via qRT-PCR. Fold induction was calculated to sham-operated control (D). (E) Chimeric CD11c-DTR mice (n = 4 each group) were used for DC depletion (25 ng/g first 2 weeks, 10 ng/g last 2 weeks) and simultaneous liver fibrosis induction using CCl4 (0.5 μL/g, three times per week). Fibrosis was determined via sirius red staining. (F) For pDC depletion, C57Bl/6 mice (n = 4) were injected every 48 hours with 120G8 antibody (500 μg/mouse) or isotype control during the last 2 weeks of CCl4-induced liver fibrosis (0.5 μL/g, three times per week for 4 weeks). Fibrosis was determined via sirius red staining. *P < 0.05. n.s., not significant.


Hepatic fibrogenesis involves multiple resident and recruited cell populations. HSCs represent the center component of this wound healing response, but other populations, including macrophages, are known positive modulators of fibrogenesis. Here, we uncover a novel function of macrophages, the promotion of HSC/myofibroblast survival. A second novel finding of our study lies in the discovery that DCs do not contribute to liver fibrosis.

Employing microarray and pathway analysis, we discovered that NF-κB, the best-characterized antiapoptotic signaling pathway[29, 30] and an important regulator of liver injury and fibrosis,[31] was a key pathway activated in HSCs by HMs. The relevance and physiologic nature of the employed in vitro coculture system is validated by the finding that this system achieves HSC gene expression patterns highly similar to those found in in vivo–activated HSCs, and that all gene expression changes and functional consequences of NF-κB activation were confirmed in vivo. Activation of the NF-κB pathway was further established by reporter assays, western blots, immunofluorescence, and qRT-PCR. Most importantly, NF-κB was activated in HSCs from fibrotic livers, and macrophage depletion reduced NF-κB activation in HSCs. The activation of NF-κB in HSCs in liver fibrosis is consistent with a previous study, but points toward macrophages instead of angiotensin II as the main trigger of NF-κB activation in HSCs.[32] Surprisingly, coculture with macrophages and macrophage-secreted cytokines such as IL-1β and TNFα did not promote HSC activation, and is consistent with the reported minor or insignificant inductions of α-SMA and Col1a1 mRNA,[33] and absence of increased α-SMA protein expression in most studies that cocultured human and murine HSCs with macrophages.[33, 34] Only one previous study found a profound and significant activation of rat HSCs by HMs.[35] In our study, macrophage-induced NF-κB activation rendered activated HSCs more resistant to cell death in vitro and in vivo, thereby promoting the persistence of activated HSCs and fibrosis. Although the rate of 1% HSC apoptosis in fibrotic livers appeared low, it reflects the rapid removal of apoptotic cells in vivo (as opposed to their accumulation in vitro), and is virtually identical to peak apoptosis rates reported by Iredale et al.[22] Thus, the observed increase to 5% HSC apoptosis is biologically highly significant, reducing the number activated myofibroblasts and limiting fibrogenic responses as reported.[11, 22, 32, 36] It is likely that increased NF-κB activation protects activated HSCs from both intrinsic and extrinsic inducers of cell death. Accordingly, our study also found that HMs induce the expression of Trail decoy receptors in HSCs in an NF-κB–dependent manner. This finding is of interest because natural killer cells, which are particularly enriched in the liver and activated during liver injury, contribute significantly to the killing of activated HSCs during liver fibrosis in a Trail-dependent manner.[11, 37, 38]

Our study identified IL-1 and TNF as main factors of HM-mediated NF-κB activation and cytoprotection in HSCs. Notably, we observed no effect of IL-1β or TNFα on HSC activation. The key role of HM-derived IL-1 and TNF in NF-κB activation and protection from HSC death was found not only in vitro but also in vivo, as demonstrated by the profound decrease in NF-κB–responsive genes in unplated, ultrapure HSC isolates from TNFR1/IL1R1 dko mice, and increased apoptosis of desmin-positive cells in TNFR1/IL1R1 dko livers after BDL. Previous studies have demonstrated reduced fibrogenesis in mice deficient in TNFR1 or IL1-R.[39, 40] In contrast to these studies, we could not observe reduced liver fibrosis in IL-1R knockout mice in three different models of liver fibrosis. This is consistent with the notion that both TNFα and IL-1β are powerful NF-κB activators, that they can likely functionally substitute each other.

Our study employed F4/80-positive HM from bile duct-ligated livers for co-culture experiments and therefore exposed HSCs to a mixture of resident and recruited macrophages typically found during fibrogenesis. Accordingly, our data show that these HM have a mixed M1/M2 phenotype as previously reported.[20] Based on our observations that converting HM into M1 phenotype increased, and into M2 phenotype reduced their ability to induce NF-κB-dependent gene expression in HSCs, we conclude that the inflammatory/M1 HM subpopulation contributes to NF-κB activation and HSC survival. It should be emphasized that the M1/M2 classification does not fully account for diverse and often overlapping biological functions of macrophage populations, particularly in the liver.[20] It is conceivable that different HM populations collaborate for the induction fibrosis in vivo, with inflammatory M1-type HM promoting HSC survival and M2-type HM affecting HSCs through other pathways. We did not find a significant impact of Gr1 status in HM on NF-κB activation in HSCs (data not shown), suggesting that both recruited and resident macrophages are capable of promoting NF-κB activation in HSCs. Clodronate did not affect HSC activation directly, nor did it alter NF-κB activation in HSCs. Moreover, our results employing DC depletion additionally excluded DC as potential contributors to clodronate effects, as we did not see a contribution of this cell type to liver fibrosis.

DCs are key regulators of inflammation and the cytokine milieu in the fibrotic liver.[12] Moreover, DCs contribute to the regression of liver fibrosis through an MMP9-dependent mechanism.[16] However, the contribution of DCs to fibrogenesis is unknown. Although we found that CD11c-positive DCs induce a moderate degree of NF-κB activation in HSCs via TNF and IL-1 production, we did not observe a role for pDC or cDC in promoting liver fibrosis in BDL- and CCl4-induced liver fibrosis. Most likely, the much lower number of DCs in the liver in comparison to HMs and the lower potency of NF-κB activation by DCs renders the contribution of DC-derived TNF and IL-1 to the overall pool and NF-κB–mediated HSC survival insignificant. In this regard, the ratio of DCs to HSCs in our coculture experiments is at least one or two magnitudes higher than the ratio that can be achieved in a fibrotic liver. Another possible explanation may be the critical role of DCs in NK cell activation, cells with well-established antifibrogenic potential.[11, 41] None of the available CD11c-DTR based ablation strategies can achieve a completely selective depletion of cDCs without affecting the composition of other immune cells.[26, 27] Even recent transgenic mouse models that avoid early neutrophilia after DC depletion still lead to neutrophilia after 2 days.[27] Although neutrophilia represents a confounding factor, we consider it unlikely that neutrophilia affects fibrogenesis based on previous studies that did not show effects on liver fibrosis.[28] These data suggest that DCs interact with and regulate other immune cells, or that increased granulopoiesis after ablation causes these secondary effects. Nonetheless, the ablation strategies employed in our study (1) avoid common side effects described for DC ablation[26] due to our use of CD11c-DTR chimeric mice and 120G8 antibody, (2) address the role of both cDC and pDC, and (3) investigate the role of cDCs in two common models of liver fibrosis.

Our study contains several limitations. First, because we observed more than 1,400 HM-regulated genes, it is likely that genes besides NF-κB–regulated genes affect HSC responses. Further studies are required to unravel the relevance of NF-κB–independent genes and pathways regulated by HM. These may include additional mediators secreted from HMs such as IL-6 and transforming growth factor β.[35, 42] Accordingly, our IPA analysis revealed Stat1/3/5 as an HM-activated pathway. Second, our studies were performed in mouse models, and further studies are required to determine whether HM-induced NF-κB activation plays a role in human fibrogenesis. As patients develop fibrosis slowly over decades, pathways that promote long-term myofibroblast survival may be particularly relevant. IL-1 and TNF inhibitors may be considered for antifibrotic therapies but may cause severe side effects. In conjunction with previous studies,[32, 43] our data support the concept that targeting the NF-κB pathway in HSCs and subsequent induction of HSC apoptosis may be a more suitable antifibrogenic strategy.

In conclusion, our study shows that HMs provide a novel link between inflammation, HSC survival, and liver fibrosis and suggests that inflammatory signaling pathways may provide additional targets for antifibrotic therapies in the liver. Future studies are needed to determine whether macrophage-mediated promotion of myofibroblast survival also promotes fibrosis in other organs.