Hepatic activation of IKK/NFκB signaling induces liver fibrosis via macrophage-mediated chronic inflammation


  • Potential conflict of interest: Nothing to report.

  • Strnad and Wirth equally contributed for the current work


Liver damage in humans is induced by various insults including alcohol abuse, hepatitis B/C virus infection, autoimmune or metabolic disorders and, when persistent, leads to development of liver fibrosis. Because the nuclear factor-κB (NF-κB) system is activated in response to several of these stresses, we hypothesized that NF-κB activation in hepatocytes may contribute to fibrosis development. To activate the NF-κB signaling pathway in a time- and cell-type-specific manner in the liver, we crossed transgenic mice carrying the tetracycline-responsive transactivator under the control of the liver activator protein promotor with transgenic mice carrying a constitutively active form of the Ikbkb gene (IKK2 protein [CAIKK2]). Double-transgenic mice displayed doxycycline-regulated CAIKK2 expression in hepatocytes. Removal of doxycycline at birth led to activation of NF-κB signaling, moderate liver damage, recruitment of inflammatory cells, hepatocyte proliferation, and ultimately to spontaneous liver fibrosis development. Microarray analysis revealed prominent up-regulation of chemokines and chemokine receptors and this induction was rapidly reversed after switching off the CAIKK2 expression. Turning off the transgene expression for 3 weeks reversed stellate cell activation but did not diminish liver fibrosis. The elimination of macrophages by clodronate-liposomes attenuated NF-κB-induced liver fibrosis in a liver-injury-independent manner. Conclusion: Our results revealed that hepatic activation of IKK/NF-κB is sufficient to induce liver fibrosis by way of macrophage-mediated chronic inflammation. Therefore, agents controlling the hepatic NF-κB system represent attractive therapeutic tools to prevent fibrosis development in multiple chronic liver diseases. (HEPATOLOGY 2012;56:1117–1128)

Liver fibrosis is the sequel of many types of chronic liver diseases including alcohol abuse, hepatitis B/C virus infection, or nonalcoholic steatohepatitis (NASH). It is characterized by a transformation of hepatic stellate cells (HSCs) from a quiescent, fat-storing phenotype into activated, extracellular matrix (ECM)-producing, alpha-smooth muscle actin (α-SMA)-positive myofibroblasts. The signals involved in this process include transforming growth factor (TGF)-β and platelet-derived growth factor (PDGF). If the injury persists, enhanced deposition of ECM proteins (e.g., collagen) results in a gradual substitution of liver parenchyma, distortion of hepatic architecture, and development of endstage liver fibrosis, which is termed liver cirrhosis. Although the presence of liver cirrhosis is well-tolerated in some patients, it often progresses to liver failure or leads to development of hepatocellular carcinoma and thus is the major cause of liver-related mortality.1

Although the process of stellate cell activation is fairly well-characterized in vitro, the in vivo situation is more complex and involves a crosstalk between different liver cell populations. Hepatocytes, the parenchymal cells of the liver, are the primary target of most human liver-damaging agents. They can activate stellate cells either directly or by way of recruitment of inflammatory cells. It has been shown that cytochrome P450 2E1 expressing hepatocytes can activate HSCs in direct coculture experiments.2 Similar effects have been observed when media from hepatocytes incubated with ferric nitrilotriacetate complex or from hepatitis C virus-replicating hepatocytes were used.3, 4 On the other hand, inflammatory cells, such as circulating monocytes, were often shown to contribute to the progression of liver fibrosis. Monocyte-derived macrophages purified from carbon tetrachloride-treated animals, or CD14+CD16+ monocytes from patients with chronic liver disease can directly activate stellate cells.5, 6 Therefore, further studies are needed to delineate the relationship between hepatocytes, inflammatory cells, and HSCs during liver fibrogenesis.

A potential key factor of hepatocyte-driven liver fibrosis could be the activation of the proinflammatory nuclear factor-κB (NF-κB) pathway in hepatocytes. The family of NF-κB transcription factors belongs to the key regulators of inflammatory processes.7 Dysregulation of NF-κB can lead to constitutive overproduction of proinflammatory cytokines, which is associated with a number of chronic inflammatory disorders.8 Constitutive activation of NF-κB is also observed in patients with liver diseases such as hepatitis B, hepatitis C, or hepatocellular carcinoma.9 However, although previous work has found that active NF-κB is associated with fibrosis,10 the exact contribution to disease development and progression remained enigmatic. In addition, the question whether activation of NF-κB is protective or disease aggravating is unresolved. In resting cells, NF-κB is localized in the cytoplasm associated with inhibitory proteins (IκB). A variety of stimuli can activate the NF-κB signaling pathway. This leads to phosphorylation, polyubiquitination, and proteasome-dependent degradation of IκB proteins. Liberated NF-κB dimers can translocate into the nucleus and regulate NF-κB-dependent gene expression.11

The IκB kinase (IKK) complex is the master regulator for activation of the NF-κB signaling pathway.12 The kinase complex comprises the two catalytic subunits, IKK1 (IKKα) and IKK2 (IKKβ), and the regulatory subunit NEMO (IKKγ), which mediates NF-κB activation in response to a number of different stimuli by phosphorylating IκB proteins.12 Genetic studies revealed that NF-κB p65 (RelA), IKK2, or NEMO have a critical role in protecting hepatocytes during an embryonic phase.13 However, in the adult liver NF-κB inhibition in hepatocytes by conditional knockout of Rela or Ikbkb (encoding IKK2), or overexpression of IκBα super-repressor has no spontaneous liver phenotype.14-16

In this study we show that hepatic activation of NF-κB signaling is sufficient to induce liver fibrosis. Activation of the NF-κB pathway leads to development of chronic inflammation, which precedes the development of liver fibrosis. Continuous NF-κB activation is necessary for the maintenance of chronic inflammation, because turning off the IKK2 overexpression leads to a rapid decrease in multiple inflammatory cytokines and later on in a decrease in activated HSCs. Development of liver fibrosis is at least in part mediated by inflammatory immune cells, given that the elimination of macrophages attenuates the liver fibrosis development. Our data show that chronic hepatocellular NF-κB activation is sufficient for liver fibrosis development by way of recruitment and activation of macrophages.


α-SMA, alpha-smooth muscle actin; ALT, alanine amino transferase; AST, aspartate amino transferase; CA, constitutively active; CT, control; DOX, doxycycline; ECM, extracellular matrix; ELISA, enzyme-linked immunosorbent assay; EMSA, electrophoretic mobility shift assay; HPF, high-power field; HPRT, hypoxanthine-guanine phosphoribosyltransferase gene; HSC, hepatic stellate cell; IKK, IκB kinase; LAP, liver activator protein; LPS, lipopolysaccharide; NASH, nonalcoholic steatohepatitis; NF-κB, nuclear factor-κB; PDGF, platelet-derived growth factor; RLU, relative light unit; SAA, serum amyloid A; TGF, transforming growth factor; tTA, tetracycline transactivator.

Materials and Methods

Animal Studies.

We crossed mice carrying a constitutively active human IKK2 (CAIKK2) allele17 under the control of a tissue-specific tetracycline-inducible system with animals expressing tetracycline-responsive transactivator (tTA) under the control of the rat liver activator protein (LAP) promotor.14

The generated mice were on a C57BL/6 and NMRI mixed background and were backcrossed at least four times to a C57BL/6 background. Studies were performed on male mice kept under specific pathogen-free conditions. The experiments were approved by the State of Baden-Württemberg in Germany and the University of Ulm Animal Care Committee.

To avoid the embryonic activation of the IKK2/NF-κB system, all mice received 0.1 g/L doxycycline in drinking water until birth. In some cases, 4-week-old animals were readministered 0.1 g/L doxycycline (DOX) in drinking water for 3 days, or 12-week-old animals were readministered DOX for 3 weeks, to study whether a continuous CAIKK2 expression is required for the observed liver phenotype. Of note, CAIKK2 mice contain a bidirectional promoter, whose activation leads to simultaneous production of both IKK2 and Photinus pyralis luciferase.17

Mice were sacrificed by way of CO2 inhalation and blood was collected from vena cava inferior. After brief centrifugation, serum was collected and used for measurement of alanine and aspartate aminotransferase levels (ALT and AST; Reflotron system, Roche). Livers were removed, weighed, and divided into pieces that were fixed in 10% formaldehyde for histological/immunohistochemical analysis, snap-frozen in liquid nitrogen for molecular or biochemical analysis, or rapidly frozen for immunofluorescence staining.

Western Blotting and Electrophoretic Mobility Shift Assay (EMSA).

For preparing whole liver extract, frozen livers were lysed in Dignam C buffer18 or in RIPA buffer.19 The lysate was centrifuged at 20,000g for 30 minutes at 4°C and the supernatant was recovered. Protein extracts were electrophoresed and subsequently blotted. Blots were incubated with the following antibodies: IKK2 (ab32135) and β-tubulin (ab6046) from Abcam, p65 (sc-372) from Santa Cruz Biotechnology, cleaved caspase 3 (9661), Parp-1 (9542), phospho-JNK (9255), JNK (9258) from Cell Signaling, keratin 18 (55148A) from Anaspec, α-SMA (A2547) from Sigma. EMSA experiments were performed as described.20

In Vivo Imaging System and In Vitro Luciferase Assay.

To monitor transgene expression, mice were anesthetized and injected intraperitoneally with 25 mM Luciferin (Synchem OHG) (150 μg/g body weight). Bioluminescence was monitored 1 minute after injection by the IVIS Imaging system 200 (Caliper Life Science). For in vitro luciferase assay, protein extract was incubated with luciferase buffer (20 mM Tris, 1.07 mM magnesium carbonate, 2.7 mM magnesium sulfate, 0.1 mM dimethyl sulfoxide [DMSO], 60 mM dithiothreitol [DTT], 1.06 mM adenosine triphosphate [ATP], 0.54 mM coenzyme A [CoA], 1 mM Luciferin) and luciferase activity was measured by Lumat LB 9507 (Berthold Technologies).

Immunofluorescence and Immunohistochemistry.

For IKK2 immunofluorescence, 4-μm-thick frozen sections were used. Slides were fixed with 4% paraformaldehyde. Slides were blocked with 5% bovine serum albumin (BSA), then incubated with antibody against IKK2, and further incubated with Alexa Fluor 488 antibody (A21206, Invitrogen). Nuclear staining was achieved by 4′,6-diamidino-2-phenylindol (DAPI). Immunohistochemical analyses for p65, Ki-67, F4/80, cleaved caspase-3, and CD3 staining were performed with 2-μm sections from paraffin-embedded samples (frozen section for CD3). Sections were deparaffinized and hydrated through graded ethanol and cooked in 10 mM citrate buffer pH 6.0 for antigen retrieval. Sections were then incubated with corresponding primary antibody. For the F4/80 immunohistochemistry, slides were treated with 3% H2O2 and blocked with 5% goat serum prior to incubation with the antibody. For cleaved caspase-3 staining, sections were blocked with 10% goat serum with 1% BSA prior to incubation with primary antibody. After incubation with secondary antibody (Dako/Jackson ImmunoResearch), slides were developed with AEC or Permanent Red systems (Dako). Experiments were performed with the following antibodies: IKK2 (ab32135, Abcam), p65 (RB1638, Neomarkers), F4/80 (ab6640, Abcam), CD3 (500A2, BD Bioscience), Ki67 (Sp6, Neomarkers), cleaved caspase-3 (ab13847, Abcam). Detailed protocols for immunofluorescence or immunohistochemistry with each antibody are available on request.

Gene Expression Analysis.

RNA was extracted from liver samples kept in RNAlater (Qiagen) by RNAeasy Mini Kit (Qiagen), and complementary DNA was generated from 2 μg of RNA using MMLV reverse transcriptase (Promega) according to the manufacturer's instructions. Quantitative real-time polymerase chain reaction (PCR) was carried out using qPCR master mix and corresponding universal probe library on Roche LC480 light cycler system (Roche). Primers are listed in Supporting Table 3. For gene expression array analysis, GeneChip Mouse Gene 1.0 ST array was used (Affymetrix). A detailed protocol for microarray experiments is provided in the Supporting Materials.

Enzyme-Linked Immunosorbent Assay (ELISA).

The serum level of serum amyloid A (SAA) was quantified with the Mouse SAA ELISA test kit (3400-1, Life Diagnostics). Experiments were performed as described by the manufacturer.

Cytometric Bead Array.

For quantification of serum levels of CCL2 and CCL5, Mouse MCP-1 Flex Set and Mouse RANTES Flex Set (BD) were used. Instrument set-up and experiments were performed with Mouse/Rat Soluble Protein Master Buffer Kit (BD) on the BD FACSArray bioanalyzer software. Data analysis was done on BD FCAP Array software. All experimental procedures were done as recommended by the manufacturer.

Hydroxyproline Assay.

Hepatic collagen levels were quantified by way of determination of hydroxyproline content as described.21 Briefly, liver samples were homogenized in distilled water. The homogenates were hydrolyzed in 6 N HCl (final concentration) by incubating at 110°C for 18 hours. The hydrolysates were filtered (Millex-HV, Millipore) and evaporated by speed vacuum centrifugation. The sediments or 10-100 μg of standards (high-purity trans-4-hydroxy-L-proline, Sigma-Aldrich) were dissolved in 50 μL of distilled water, then mixed with 450 μL of 56 mM chloramines-T (Sigma-Aldrich) in acetate-citrate buffer (pH 6.5) and incubated for 25 minutes at room temperature. Subsequently, 500 μL of Ehrlich's solution (Fluka) was added, mixed, and incubated at 65°C for 20 minutes, followed by reading the absorbance at 562 nm.

Macrophage Elimination by Clodronate.

Control liposome and clodronate liposome were synthesized as described22 and injected intraperitoneally (10 μL/g mouse) from the age of 3 weeks on. At the age of 12 weeks all animals were sacrificed and analyzed further.

Statistical Analysis.

Data are shown as means ± standard deviation (SD). Statistical significance was determined using a two-tailed Student's t test. P < 0.05 was considered significant.


Hepatocellular NF-κB Activation Induces the Development of Liver Fibrosis.

To understand the effect of NF-κB activation in the liver, we crossed mice carrying a constitutively active IKK2 (CAIKK2) allele17 under the control of a tetracycline-regulated promoter with animals expressing tTA under the control of the LAP promotor.14 The resulting double transgenic mice were termed CAIKK2LAP (Supporting Fig. 1A). Given the known critical role of NF-κB in hepatocytes during embryonic development, we repressed transgenic IKK2 expression by DOX administration in the drinking water to the pregnant mothers. Using this strategy, double transgenic mice were born at the expected Mendelian frequency. Measurements of luciferase, which was used as a reporter gene, confirmed the absence of transgene expression in the DOX-administered animals (Supporting Fig. 2A). Removal of DOX at birth led to induction of luciferase, whereas no luciferase activity was observed in animals continuously treated with DOX (Supporting Fig. 1B). In vivo imaging and luciferase assays revealed that expression of the transgene reporter luciferase was restricted to the liver both in vivo and in vitro (Supporting Figs. 1B, 2A). Furthermore, expression of CAIKK2 transgene was detectable in the liver, but not in isolated HSCs or Kupffer cells (Supporting Fig. 1C). Immunofluorescence staining and western blot analyses demonstrated CAIKK2 protein expression in the liver (Supporting Fig. 2B,C), whereas no CAIKK2 expression was observed in control mouse livers and CAIKK2LAP mouse livers from DOX-treated animals (Supporting Fig. 2C). CAIKK2 expression led to constitutive activation of the NF-κB signaling pathway, as evidenced by the increased NF-κB DNA-binding activity in EMSA assay (Supporting Fig. 2D) and the nuclear accumulation of NF-κB/p65 in hepatocytes (Supporting Fig. 2E). There was no NF-κB/p65 nuclear accumulation when CAIKK2LAP mice were kept under DOX (Supporting Fig. 2E), confirming again the tight regulation of the transgenic system.

Postnatal NF-κB activation in CAIKK2LAP mice did not result in any lethality, an obvious growth defect, or clinical signs of liver failure (body weight: 4-week-old, control 14.2 ± 3.1 g, CAIKK2LAP 14.3 ± 4.6 g, P = 0.9; 12-week-old, control 26.0 ± 2.3 g, CAIKK2LAP 26.5 ± 2.2 g, P = 0.6; Supporting Fig. 1D and data not shown). Furthermore, there was no significant difference in liver weight (P = 0.9) and liver weight/body weight ratio (P = 0.7) between 4-week-old control animals and CAIKK2LAP mice (Fig. 1A). However, the livers from 12-week-old CAIKK2LAP animals were macroscopically distinguishable by their marked enlargement (liver weight, control 1.3 ± 0.1 g, CAIKK2LAP 1.9 ± 0.5 g, P = 6 × 10−4; liver weight/body weight ratio, control 0.05 ± 0.003, CAIKK2LAP 0.07 ± 0.02, P = 4 × 10−4), paleness, and rigidity compared to livers from control littermates (Fig. 1A,B). Histological analyses revealed that the livers from 12-week-old CAIKK2LAP mice exhibited mononuclear leukocytic infiltration of the portal tracts and a predominantly diffuse inflammation of the lobular parenchyma associated with a variable extent of hepatocellular damage (Desmet score: control 0.2 ± 0.4, CAIKK2LAP 1.7 ± 1.2, P = 1 × 10−4; Fig. 1C,D). Portal and intralobular inflammation was also present in 4-week-old transgenic, but not in nontransgenic animals (Desmet score: control 0, CAIKK2LAP 2.5 ± 0.8, P = 7 × 10−6; Fig. 1C,D). In addition, both 4-week- and 12-week-old CAIKK2LAP mice presented with mildly elevated ALT (4-week-old, control 18 ± 3, CAIKK2LAP 40 ± 19 IU/L, P = 2 × 10−3; 12-week-old, control 22 ± 11 IU/L, CAIKK2LAP 44 ± 15 IU/L, P = 9 × 10−4) and AST levels (4-week-old, control 45 ± 9 IU/L, CAIKK2LAP 88 ± 30 IU/L, P = 1 × 10−4; 12-week-old, control 46 ± 17 IU/L, CAIKK2LAP 91 ± 38 IU/L, P = 2 × 10−3), which reflects the rather modest extent of liver injury (Fig. 1E). The extent of hepatic inflammation as well as ALT/AST levels did not differ between 4- and 12-week-old CAIKK2LAP mice. Furthermore, CAIKK2LAP mice did not exhibit increased apoptosis levels as measured by cleaved caspase 3, keratin 18, and Parp-1. On the other hand, all markers were clearly elevated in lipopolysaccharide (LPS)-stimulated, TAK1-deficient animals (TAK1LPC-KO),23 which serves as a model of a loss of protective hepatocellular NF-κB signaling (Supporting Fig. 3A). Interestingly, livers from CAIKK2LAP mice showed hyperproliferation as demonstrated by Ki67 staining and expression of the G1-S-phase transition marker Cyclin E. The extent of hepatic proliferation was comparable to TAK1LPC-KO mouse livers or livers from partially hepatectomized mice (Supporting Fig. 3B). The observed hyperproliferation in CAIKK2LAP mice could be due to activation of the cell cycle driven by the NF-κB signaling, or by subordinately activated JNK signaling (Supporting Fig. 3C).

Figure 1.

Expression of constitutively active IKK2 isoform leads to mild chronic hepatitis. (A) Liver weight and liver weight/body weight ratio from 4-week-old and 12-week-old mice (n = 11). (B) Macroscopic appearance of livers of a representative 12-week-old control and CAIKK2LAP mouse demonstrate the enlargement, paleness, and irregular appearance of the transgenic livers (scale bar = 1 cm). (C) Representative hematoxylin and eosin (H&E) staining from control and CAIKK2LAP mouse livers (scale bar = 100 μm). (D) Inflammation was scored in the H&E sections according to the Desmet method. Analysis was done with sections from 4-week-old (control [CT] n = 7, CAIKK2LAP n = 6) and 12-week-old animals (control n = 13, CAIKK2LAP n = 17). (E) ALT and AST levels in the 4- and 12-week-old mice (n = 11).

Sirius-red staining revealed a hepatic fibrosis in 12-week-old CAIKK2LAP mice (Fig. 2A), whereas no significant fibrosis was seen in 4-week-old CAIKK2LAP mice nor in nontransgenic animals at any age (Fig. 2A; Supporting Fig. 4A). The extent of fibrosis in 12-week-old CAIKK2LAP mice was variable, ranging from mild portal fibrosis (Desmet score 1) to bona fide cirrhosis (Desmet score 4, seen in 2 out of 16 analyzed animals) (control 0.06 ± 0.3, CAIKK2LAP 1.9 ± 1.3, P = 3 × 10−6; Fig. 2B and data not shown). These data were further confirmed by elevated hydroxyproline content (biochemical marker of collagen deposition; control 1 ± 0.1, CAIKK2LAP 1.8 ± 0.1, P = 3 × 10−8), morphometrical analysis of Sirius-red stained sections, as well as increased hepatic collagen messenger RNA (mRNA) levels (gene Col1a1) in CAIKK2LAP versus control mice (relative to hypoxanthine-guanine phosphoribosyltransferase gene [HPRT], control 0.04 ± 0.03, CAIKK2LAP 3.0 ± 2.0, P = 0.002; Fig. 2A,D,E). The higher collagen deposition in CAIKK2LAP mice was likely due to increased HSC activation, given that α-SMA (gene Acta2), an established HSC activation marker, was significantly elevated, both at the mRNA and protein levels (Fig. 2C,E). In addition, we also observed higher levels of several fibrosis-associated genes such as Tgfb1 and Icam (Fig. 2E). Of note, elevated Col1a1 and Acta levels were already seen in 4-week-old CAIKK2LAP mouse livers (Supporting Fig. 4B), suggesting that these mice already display a significant HSC activation, but no appreciable collagen deposition.

Figure 2.

Expression of constitutively active IKK2 isoform leads to spontaneous liver fibrosis development. (A) Sirius-red staining with morphometric analysis of Sirius-red-positive area (CT n = 13 CAIKK2LAP n = 16) visualizes the spontaneous liver fibrosis development observed in 12-week-old CAIKK2LAP mice, but not in control animals (scale bar = 200 μm). (B) Fibrosis status from Sirius-red-stained sections was scored according to the Desmet-Scheuer scoring system (n = 16). (C) Expression of α-SMA, an established stellate cell activation marker, was examined by western blotting in 12-week-old control (CT) and CAIKK2LAP mice. α-SMA protein was strongly expressed in CAIKK2LAP mouse livers compared to control animal livers. NF-κB/p65 served as loading control. (D) Collagen deposition in the liver was quantified by hydroxyproline content measurement. Increased hydroxyproline content was observed in liver extracts from 12-week-old CAIKK2LAP mice as compared to those from control animals (CT) (n = 8). Data are shown as ratio to control mice, which was arbitrarily set as 1. (E) Up-regulation of fibrosis-associated genes in the livers from CAIKK2LAP mice. Results from quantitative real-time PCR experiments are shown. Bars represent the ratio of gene to HPRT expression (n = 7).

Hepatocellular NF-κB Activation Leads to Liver Fibrosis Development by Way of Recruitment of Inflammatory Cells.

To study the pathogenesis of liver fibrosis development in CAIKK2LAP mice, we performed microarray analyses using livers from 4-week-old mice. In all, 1,043 genes were significantly overexpressed in double-transgenic animals compared to controls (Supporting Table 1). Gene ontology analysis revealed hepatocyte stress reaction, inflammation, and chemotaxis as the major pathways altered in CAIKK2LAP animals (Supporting Table 2). The altered expression of selected genes (SAA isoforms Saa1, Saa2, and Saa3; chemokines Ccl2, Ccl5, Cxcl2, Ccl20, and Cxcl10; chemokine receptors Ccr2 and Cxcr4) and the up-regulation of macrophage-related Tnfa, Il6, and Mmp9 was confirmed by quantitative real-time PCR (Fig. 3). Furthermore, elevated SAA, CCL2, and CCL5 serum levels were observed in CAIKK2LAP mice as compared to controls (Fig. 3). To further characterize the hepatic inflammation in CAIKK2LAP mice, we performed immunohistochemical stainings. These revealed a mixed population of cells including abundant macrophages, T lymphocytes, as well as neutrophils (Fig. 4).

Figure 3.

Presence of constitutively active IKK2 isoform results in profound changes in hepatic expression profile. Microarray results were corroborated by quantitative real-time PCR experiments (4-week-old mice). Hepatic expression of stress response genes, several chemokines, and chemokine receptors, as well as macrophage-associated factors Tnfa, Il6, and Mmp9, was up-regulated in the livers from CAIKK2LAP mice as compared with control animals (CT). Results are normalized to HPRT expression (n = 6) (for Tnfa, Il6, and Mmp9 4-week-old n = 8, for Mmp9 12-week-old n = 6 [CT] or n = 9 [CAIKK2LAP]). To quantify protein level in serum, we performed ELISA for SAA and cytometric bead arrays for CCL2 and CCL5 measurement. Experiments were performed with serum from 4-week-old control mice (CT) (n = 15) and CAIKK2LAP mice (n = 11).

Figure 4.

Hepatic activation of NF-κB signaling induces liver inflammation. (A,B) Immunohistochemical characterization of inflammatory infiltrates. Tissue sections from control (CT) and CAIKK2LAP mice were immunostained with antibodies recognizes F4/80 for macrophages (A) or CD3 for T lymphocytes (B) (scale bar = 200 μm). Strong infiltration of macrophages and T lymphocytes were observed in the liver sections from 4-week-old CAIKK2LAP mice and was confirmed by morphometric quantification (n = 5 for F4/80-staining, 3 control and 4 CAIKK2LAP samples for CD3-staining), whereas only few immune cells were detected in control mouse livers. (C) Morphometric quantification of Gr-1-positive neutrophils in the liver sections from 4-week-old control (CT) and CAIKK2LAP mice. (D) Hepatic mRNA levels for genes indicating macrophage infiltration/activation were measured in 4-week-old control mice and CAIKK2LAP mice by quantitative real-time PCR. Several activation markers for macrophages are up-regulated in the liver from CAIKK2LAP mice as compared to those from control mice (CT). Bars represent the ratio of gene to HPRT expression (n = 6).

To test whether continuous NF-κB activation is needed for the observed changes, we compared gene expression in 4-week-old mice with sustained NF-κB activation with animals where CAIKK2 expression was inhibited for 3 days by DOX readministration. DOX readministration for 3 days inhibited CAIKK2 expression (Fig. 5A). Histological analyses revealed that the livers from DOX-readministered mice exhibited less inflammation (Desmet score: control 0, CAIKK2LAP 0.8 ± 0.5) compared to those from mice without DOX readministration (Desmet score: control 0, CAIKK2LAP 2.5 ± 0.8, P = 6 × 10−3) (Fig. 5B,C). Expression of stress response genes and chemokines was reversed in DOX-treated animals, suggesting that these processes require continuous NF-κB activation (Fig. 5D).

Figure 5.

Continuous activation of NF-κB is required for sustained expression of stress response genes and chemokines as well as for sustained hepatic stellate cell activation. (A) Expression of CAIKK2 was determined by protein immunoblot using antibodies against IKK2 and NF-κB/p65. The latter was used as a loading control. IKK2 was highly expressed in CAIKK2LAP mice kept without doxycycline, whereas no detectable signal was seen in control animals (CT) or transgenic mice readministered DOX for 3 days. (B) Extinction of CAIKK2LAP expression reverses the inflammatory phenotype. H&E-stained liver sections from control (CT), 4-week-old CAIKK2LAP mice, and 4-week-old control and CAIKK2LAP mice readministered DOX for 3 days are shown (scale bar = 100 μm). (C) Inflammation was quantified according to Desmet. Analysis was done with H&E-stained sections from 4-week-old animals kept without DOX (control n = 7, CAIKK2LAP n = 6), or with readministered DOX for 3 days (control (CT) n = 6, CAIKK2LAP n = 4). After 3 days DOX readministration the livers from CAIKK2LAP animals exhibited a reduced inflammation. (D) To see whether continuous activation of NF-κB is needed for sustained expression of stress response genes and chemokines, expression of indicated genes in the livers from 4-week-old mice was evaluated by quantitative real-time PCR. No DOX indicates that animals were not treated with DOX after birth (n = 6), 3 days DOX indicates that the animals were readministered DOX for 3 days at the age of 4 weeks (control [CT] n = 5, CAIKK2LAP n = 4). Bars represent the ratio of gene to HPRT expression. (E) Continuous activation of NF-κB is required for sustained hepatic stellate cell activation. No DOX indicates that the mice were not readministered DOX after birth, whereas “3 weeks DOX” indicates that the mice at the age of 12 weeks were readministered DOX for 3 weeks. Note that silencing of transgenic CAIKK2 by way of addition of DOX for 3 weeks was sufficient to reverse HSC activation. Quantitative real-time PCR was used to measure the liver expression of selected genes compared to HPRT expression (n = 6).

Furthermore, 12-week-old animals were readministered DOX for 3 weeks. Histological analyses, as well as hydroxyproline assay, revealed that despite the NF-κB system being switched off, hepatic fibrosis did not regress within this time frame (Supporting Fig. S5). Although 3 weeks might have been too short to significantly reduce the extent of liver fibrosis, HSC activation markers were all down-regulated after readministration of DOX (Fig. 5E), suggesting that activation of HSCs is dependent on sustained activation of the hepatic NF-κB system.

Given the increased presence of macrophage surface markers as well as genes involved in macrophage activation in our microarray analysis (Fig. 4D), we hypothesized that activation of hepatocellular NF-κB signaling leads to liver fibrosis development by way of macrophage recruitment. To test whether hepatocyte-mediated recruitment of macrophages contributes to liver fibrogenesis, we injected clodronate liposomes, an established macrophage-depleting agent. Injection of clodronate liposomes from age 3 to 12 weeks resulted in a marked decrease in macrophage surface marker F4/80, as well as attenuated lysozyme M and Cxcl10 expression, whereas CAIKK2 expression was intact (Fig. 6A,B; Supporting Fig. S6B). Unaltered levels of F4/80 expression were observed in animals injected with control liposomes (Fig. 6B). Clodronate administration did not reduce the extent of hepatic damage (Supporting Fig. S6A), or the extent of overall inflammation as suggested by unaltered levels of several inflammation-related genes (Supporting Fig. S6B). On the other hand, we observed attenuated fibrogenesis as evidenced by Sirius-red staining (Fig. 6C) and by Desmet scoring (Fig. 6D). A reduction of collagen deposition was also confirmed by hydroxyproline assay and morphometrical analysis of Sirius-red-stained sections (Fig. 6C,E). Thus, our data indicate that sustained hepatocellular NF-κB activation leads to liver fibrosis development by way of recruitment and activation of macrophages. On the other hand, macrophage depletion was effective in significantly reducing liver fibrosis, but not sufficient to reduce the extent of liver injury or overall hepatic inflammation, which suggests that additional cell types contribute to the observed liver phenotype.

Figure 6.

Elimination of macrophages attenuates hepatic fibrosis in CAIKK2LAP mice. (A) Immunofluorescence staining of liver sections for hepatocellular CAIKK2 expression in control mice (CT), CAIKK2LAP mice injected control liposomes (CL), or clodronate liposomes (CD). Costaining with DAPI was used to label the nuclei (scale bar = 100 μm). (B) The effect of clodronate liposome injections on liver macrophage elimination was analyzed by F4/80 immunohistochemistry. Representative anti-F4/80 immunostainings from 12-week-old control mouse (CT), and CAIKK2LAP mice injected either control liposomes (CL) or clodronate liposomes (CD) are shown (scale bar = 200 μm). (C) Macrophage elimination resulted in attenuation of IKK2-induced liver fibrosis. Fibrillar collagen deposition was evaluated by Sirius-red staining as well as morphometric analysis of Sirius-red-positive area (CT n = 13, CL n = 7, CD n = 6). Representative liver sections from a 12-week-old control mouse and control liposome (CL)- or clodronate liposome (CD)-injected CAIKK2LAP mice are shown (scale bar = 200 μm). (D) Desmet-Scheuer scoring for control animals (CT) (n = 16), control liposome (CL) (n = 7), or clodronate liposome (CD) (n = 6)-injected CAIKK2LAP mice. Note that injection of clodronate liposomes attenuated CAIKK2-induced liver fibrosis. (E) Hydroxyproline content as a measurement of hepatic collagen deposition. Note that 12-week-old, control liposome-injected (CL) CAIKK2LAP mouse livers (n = 6) contain significantly more hydroxyproline than control animals (n = 8) or clodronate liposome (CD)-injected CAIKK2LAP mice (n = 6). Data are shown as ratio to control mice, which were arbitrarily set as 1.


Here we demonstrate that hepatocellular activation of IKK/NF-κB is sufficient to induce liver fibrosis within several weeks. A constitutive activation of NF-κB was reported in patients with hepatitis B or hepatitis C infection and ethanol administration was also shown to activate NF-κB signaling.9 Therefore, this model reflects liver fibrosis development in many human situations that are associated with primary affection of hepatocytes. Similar to human chronic liver diseases, hepatocellular NF-κB activation results in only mild elevation of transaminase serum levels. Thereby, it differs from other models where liver fibrosis develops as a consequence of massive hepatocellular necrosis/apoptosis, e.g., by hepatic disruption of Bcl-xL, TAK1 (an upstream kinase for the IKK complex), or Prohibitin 1.23-26 These unique properties make hepatocellular NF-κB activation an attractive model to study liver fibrosis development in human liver disorders.

The notion that sustained NF-κB activation within hepatocytes induces liver fibrosis development is also supported by findings from other tissues. For example, continuous activation of NF-κB in the pancreas (e.g., acinar cells) led to leukocyte infiltration, up-regulation of inflammatory cytokines and chemokines, and pancreatic fibrosis.27, 28 However, NF-κB represents a complex system, which cannot be viewed simply as pro- or antifibrogenic. To that end, a complete abrogation of the NF-κB signaling by hepatic ablation of NEMO or TAK1 caused the spontaneous development of liver fibrosis.23, 25, 29 In these models, liver fibrosis development is likely caused by the loss of the prosurvival NF-κB signaling that results in massive hepatocellular damage and secondary inflammatory reaction. On the other hand, hepatic IKK2 overexpression seems to rather stimulate hepatocellular stress signaling, which then leads to recruitment of inflammatory cells (Fig. 7).

Figure 7.

Modulation of NF-κB signaling leads to liver fibrosis development. NF-κB signaling represents a double-edged sword in that both its depletion and activation results in liver damage and ultimately in liver fibrosis development. The loss of hepatocellular NF-κB signaling as seen in NEMO- or TAK1-knockout mice highlights the important prosurvival function of NF-κB and leads to massive hepatocellular damage with a secondary inflammatory reaction. Conversely, activation of NF-κB signaling by way of expression of constitutively active IKK2 isoform, alcohol abuse, or hepatitis B/C infection cause only moderate hepatocellular damage, but leads to chronic liver inflammation. This hepatic inflammation requires continuous NF-κB activation (as shown in animals with stopped IKK2 expression) and contributes to liver fibrosis development in that liver fibrogenesis is attenuated in mice in whose macrophages were depleted. These data suggest that activation of NF-κB signaling seen in several human liver diseases such as viral hepatitis or alcoholic liver disease contributes to liver fibrosis development.

The molecular link between hepatocellular injury and liver fibrosis development remains a subject of intense debate. Although HSCs are the major hepatic collagen producers, the precise mode of their activation remains unresolved. Some evidence shows that hepatocytes can directly activate stellate cells,2-4 whereas others emphasize the role of inflammatory cells in this process.5, 6 Given the established role of NF-κB in regulation of both the innate and adaptive immune responses, an induction of chronic inflammation is likely responsible for liver fibrosis development in CAIKK2LAP mice. This is further supported by several of our findings, including: (1) upregulation of multiple inflammatory genes in the microarray analysis; (2) the notion that upregulation of inflammatory genes relies on continuous NF-κB activation; and (3) the fact that clodronate liposome-mediated macrophage depletion attenuates the disease. Several inflammatory genes including chemokines and chemokine receptors were previously described to be regulated by NF-κB pathway and might be responsible for the observed phenotype. For example, the chemokine CCL2 and the chemokine receptors CCR1 and CCR2 were implicated in monocyte recruitment during experimental liver fibrosis.5, 30-33

In our study, macrophage depletion attenuated experimental liver fibrosis development without affecting the extent of liver injury or the extent of overall inflammation. This suggests that not only the magnitude, but also the type of liver injury/inflammation influences liver fibrogenesis. Although macrophages were crucial for fibrosis development, the contribution of liver injury to this process needs to be investigated further.

Several other findings provide a support for the important role of macrophages in liver fibrosis development. For example, macrophage depletion can reduce carbon tetrachloride-induced liver fibrosis.34 Macrophages and also infiltrating monocytes are considered the main producers of transforming growth factor beta (TGF-β), one of the most powerful mediators of HSC activation in vitro and in vivo.35 Furthermore, macrophage-produced chemokines contribute to additional recruitment of inflammatory cells.36

In summary, our study provides an important link between hepatocellular NF-κB activation, induction of chronic inflammation, and liver fibrosis development, which might be of relevance for liver disease development in multiple chronic liver disorders.


We thank Olena Sakk and Vadim Sakk for help with establishing the transgenic model and in characterization of the fibrosis phenotype. We also thank Melanie Gerstenlauer, Kristina Diepold, Birgit Rettenmeier, and Julia Melzner for histological experiments, and Susanne Schatz for help with the mouse studies. We thank Sibille Sauer-Lehnen and Carmen G. Tag for technical assistance, and Karina Kreggenwinkel for helpful discussion. We thank Prof. Hermann Bujard for providing the LAP-tTA mice, Dr. André Lechel, and Prof. Karl Lenhard Rudolph for partial-hepatectomized mouse livers. Author contributions: Study concept and design: Y.S., P.S., T.W.; Acquisition of data: Y.S., F.L., S.G., K.F., N.G., S.E., K.H.H., N.H., A.S., S.W.; Analysis and interpretation of data: Y.S., F.L., S.G., K.F., K.H.H., N.H., A.S., P.S., T.W.; Drafting the article: Y.S., F.L., K.H.H., P.S., T.W.; Statistical analysis: Y.S., K.H.H.; Obtained funding: Y.S., P.S., T.W.; Discussion: F.K. M.S. K.S.K. S.K.; Technical or material support: S.E. T.L. B.B.