Phagocytosis of apoptotic bodies by hepatic stellate cells induces NADPH oxidase and is associated with liver fibrosis in vivo


  • Shan-Shan Zhan,

    1. Department of Internal Medicine, Division of Gastroenterology and Hepatology, Clinical Nutrition and Vascular Medicine, UC Davis Medical Center, Sacramento, CA
    Search for more papers by this author
    • Shan-Shan Zhan and Joy X. Jiang contributed equally to this study.

  • Joy X. Jiang,

    1. Department of Internal Medicine, Division of Gastroenterology and Hepatology, Clinical Nutrition and Vascular Medicine, UC Davis Medical Center, Sacramento, CA
    Search for more papers by this author
    • Shan-Shan Zhan and Joy X. Jiang contributed equally to this study.

  • Jian Wu,

    1. Department of Internal Medicine, Division of Transplant, Clinical Nutrition and Vascular Medicine, UC Davis Medical Center, Sacramento, CA
    Search for more papers by this author
  • Charles Halsted,

    1. Department of Internal Medicine, Division of Endocrinology, Clinical Nutrition and Vascular Medicine, UC Davis Medical Center, Sacramento, CA
    Search for more papers by this author
  • Scott L. Friedman,

    1. Department of Internal Medicine, Division of Liver Diseases, Mount Sinai School of Medicine, New York, NY
    Search for more papers by this author
  • Mark A. Zern,

    1. Department of Internal Medicine, Division of Transplant, Clinical Nutrition and Vascular Medicine, UC Davis Medical Center, Sacramento, CA
    Search for more papers by this author
  • Natalie J. Torok

    Corresponding author
    1. Department of Internal Medicine, Division of Gastroenterology and Hepatology, Clinical Nutrition and Vascular Medicine, UC Davis Medical Center, Sacramento, CA
    • UC Davis Medical Center, Patient Support Services Building, 4150 V Street, Suite 3500, Sacramento, CA 95817
    Search for more papers by this author
    • fax: 916-734-7908

  • Potential conflict of interest: Nothing to report.


Hepatic stellate cell activation is a main feature of liver fibrogenesis. We have previously shown that phagocytosis of apoptotic bodies by stellate cells induces procollagen α1 (I) and transforming growth factor beta (TGF-β) expression in vitro. Here we have further investigated the downstream effects of phagocytosis by studying NADPH oxidase activation and its link to procollagen α1 (I) and TGF-β1 expression in an immortalized human stellate cell line and in several models of liver fibrosis. Phagocytosis of apoptotic bodies in LX-1 cells significantly increased superoxide production both in the extracellular and intracellular milieus. By confocal microscopy of LX-1 cells, increased intracellular reactive oxygen species (ROS) were detected in the cells with intracellular apoptotic bodies, and immunohistochemistry documented translocation of the NADPH oxidase p47phox subunit to the membrane. NADPH oxidase activation resulted in upregulation of procollagen α1 (I); in contrast, TGF-β1 expression was independent of NADPH oxidase activation. This was also confirmed by using siRNA to inhibit TGF-β1 production. In addition, with EM studies we showed that phagocytosis of apoptotic bodies by stellate cells occurs in vivo. In conclusion, these data provide a mechanistic link between phagocytosis of apoptotic bodies, production of oxidative radicals, and the activation of hepatic stellate cells. (HEPATOLOGY 2006;43:435–443.)

Liver fibrosis is a wound healing process that is elicited by various toxic stimuli.1, 2 At the center of the fibrogenic process are the hepatic stellate cells (HSC), which are normally quiescent and produce only small amounts of extracellular matrix (ECM) components for the formation of the basement membrane.3 Exposure of HSC to soluble factors that include reactive oxygen species (ROS) or cytokines from damaged hepatocytes, activated Kupffer cells, or endothelial cells4 leads to a morphological and functional transition to myofibroblast-like cells.5 One of the major profibrogenic mediators is transforming growth factor-beta1 (TGF-β1); it is upregulated in activated HSC, and participates in multiple phases of HSC activation.6, 7 The pathways leading to HSC activation are complex and require better elucidation.

Apoptosis or programmed cell death is also a common feature of chronic liver disease; the end-result of ongoing toxic stimuli that result in the death of hepatocytes. Apoptosis results in the generation of apoptotic bodies (AB), which are subsequently cleared by phagocytosis. After engulfing AB, macrophages secrete TGF-β.8 Indeed, we have previously demonstrated that HSC can phagocytose AB in vitro, which triggers a profibrogenic response with upregulation of TGF-β1 and procollagen α1 (I) expression.9 This study suggested for the first time that there may be a direct link between liver injury, apoptosis of hepatocytes, phagocytosis of AB, and activation of HSC.

Phagocytosis also can activate NADPH oxidase,10 an enzyme primarily found in phagocytic cells, as well as fibroblasts and vascular smooth muscle cells,11, 12 where it is termed “nonphagocytic” NADPH oxidase (NOX). NADPH oxidase is a membrane-bound enzyme that catalyzes the production of superoxide from oxygen and NADPH. It is dormant in the resting phagocyte but becomes catalytically active when the cell is activated. During activation, the cytosolic subunits, including p47phox, migrate to the cell membrane, where they associate with membrane-bound subunits to assemble the active enzyme. The intracellular organelles then fuse with the phagocytic vesicles, resulting in the delivery of superoxide outward into the extracellular environment and inward into the phagocytic vesicle.13 HSC have recently been shown to express NADPH oxidase,14–16 and thus NADPH oxidase may become activated in HSC during phagocytosis. Therefore, NADPH oxidase could be a crucial component of the signaling cascade linking phagocytosis and fibrogenesis.

Here we show that NADPH oxidase activation occurs after phagocytosis of AB by HSC and that this appears to be an essential linking process between phagocytosis and fibrogenic activation. In addition, electron microscopy (EM) and confocal microscopic studies of livers undergoing fibrogenesis from different etiologies indicate the presence of phagocytosing HSC.


HSC, hepatic stellate cells; ROS, reactive oxygen species; TGF-β1, transforming growth factor beta1; AB, apoptotic bodies; EM, electron microscopy; BDL, bile duct ligation; FDE, folate deficient/ethanol diet; UV, ultraviolet; SOD, superoxide dismutase; DPI, diphenylene iodonium; TAMRA, carboxytetramethyl rhodamine succinimidyl ester; PBS, phosphate-buffered saline; RT-PCR, reverse transcription polymerase chain reaction; siRNA, small interfering RNA; DCFDA, 2′,7′-dichlorofluorescein diacetate.

Materials and Methods

Human Biopsy Samples.

Human liver biopsy specimens from patients with recurrent post-transplantation hepatitis C were used in the EM studies. Informed consent was obtained from each patient; specimens were obtained according to the guidelines of the approved institutional IRB protocol. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki.


For the bile duct ligation (BDL) experiments, male Sprague-Dawley rats were used (Charles River, Portage, MI). BDL was performed with the rats under pentobarbital anesthesia (60 mg/kg, intraperitoneally). The common bile duct was ligated in two locations. Animals were sacrificed 1 week after surgery, and the liver specimens were fixed and processed for EM studies. Control animals underwent sham operations. Yucatan micropigs (Sinclair Farms, Columbia, MO) on a combined folate deficient/ethanol diet with substitution of 40% of calories as ethanol17 (FDE) were also used for the EM studies. After 14 weeks on this diet, the micropigs were sacrificed, and liver samples were obtained and processed for EM. Stellate cell isolation from rats was performed according to Geerts et al.18 The animals were housed in facilities approved by the National Institute of Health. All procedures were reviewed and approved by the Animal Welfare Committee of the University of California Davis.

Cell Culture.

A human immortalized HSC line LX-1 was used.19 For positive control, a mouse macrophage-like cell line RAW 264.7 was used. To generate AB, HepG2 cells were used. LX-1, primary rat stellate cells, and RAW 264.7 were cultured in Dulbecco's minimum essential medium (Invitrogen, Carlsbad, CA) with 10% fetal bovine saline, HepG2 cells were maintained in MEM, with 10% fetal bovine serum.

Preparation of Apoptotic Bodies.

Apoptotic bodies (AB) were generated from HepG2 cells by exposing them to ultraviolet (UV) irradiation (100 mJ/cm2, 142 seconds), as described previously.9 AB were collected 48 hours after the UV irradiation; at this time 80% of cells were dead. Of this, 21% were necrotic and 79% apoptotic.

Assay for Superoxide Production as the Measure of NADPH Oxidase Activity.

To evaluate superoxide release, a superoxide dismutase (SOD)-inhibitable cytochrome c reduction assay was performed.20, 21 LX-1 cells were plated at a density of 106/mL and incubated with the AB in the presence or absence of diphenylene iodonium (DPI, 10 μmol/L, for 30 minutes) (Sigma-Aldrich, St. Louis, MO), an NADPH oxidase inhibitor.12 The supernatant was collected at 1 hour, 6 hours, and 24 hours after incubation with the AB. Alternatively, to study intracellular release of superoxide, cell lysates were obtained according to standard methods using RIPA lysis buffer (at the above time points). Cytochrome c (160 μmol/L) (Sigma-Aldrich) was added in the presence (100 U/mL) or absence of SOD (Sigma-Aldrich) and incubated for 20 minutes at 37°C. The values of superoxide-dependent reduction of cytochrome c were obtained by subtracting the absorbance readings at 550 nm of the samples without SOD from those with SOD. The amount of reduced cytochrome c was calculated,21 and the final result was reported as fold changes above the control value. To visualize superoxide production, LX-1 cells were incubated with a redox-sensitive dye 2′,7′-dichlorofluorescein diacetate (DCFDA) (20 μmol/L; Molecular Probes Inc., Eugene, OR), for 30 minutes at 37°C after phagocytosis of carboxytetramethyl rhodamine succinimidyl ester (TAMRA)-labeled apoptotic bodies. To inhibit p47phox migration, the cells were preincubated with the inhibitor ML-7 (75 μmol/L, for 30 minutes) (Calbiochem, San Diego, CA). Image analysis and co-localization of signals was performed by confocal microscopy (Zeiss, Thornwood, NY). The excitation and emission wavelengths for the DCFDA were 488 nm and 505 nm, and for TAMRA 543 nm and 580 nm, respectively.


LX-1 cells were incubated with TAMRA-labeled AB for 6 hours, and then were fixed with 4% paraformaldehyde for 20 minutes. Cells were permeabilized by 0.1% Triton-X 100 [in phosphate-buffered saline (PBS) with 1% bovine serum albumin] for 30 minutes followed by washing and then incubated with an anti-p47phox rabbit antibody (1:2000) (Upstate, NY) at 4°C, overnight. After washing with PBS, a goat anti-rabbit secondary antibody conjugated with FITC (1:300) was applied for 1 hour at room temperature. For tissue staining, 4 μm micropig liver sections were boiled in citrate buffer for 5 minutes. Slides were then incubated in 50 mmol/L NH4Cl/PBS at room temperature for 30 minutes. Blocking was done with 2% bovine serum albumin for 30 minutes, then incubation with first antibodies [E-cadherin 1:1,000, BD Transduction Laboratories (San Jose, CA)., anti-SMA 1:200, Epitomics Inc., Burlingame, CA)]. The secondary antibodies were Cy3-conjugated Affinipure Goat Anti-Mouse, 1:1000, (Jackson ImmunoResearch Lab. Inc., West Grove, PA), and Alexa Fluor 488 anti-rabbit IgG 1:1000, (Molecular Probes, Eugene, OR; Invitrogen, Carlsbad, CA). The images were analyzed by confocal microscopy.

Real-Time Polymerase Chain Reaction to Study Activation of HSC.

Real-time quantitative reverse transcription polymerase chain reaction (RT-PCR) was performed to evaluate the effects of superoxide production on the expression of TGF-β1 and procollagen α(1) I. After incubation with AB for 48 hours, LX-1 cells were harvested, total RNA extracted, using the TRIzol reagent (Invitrogen). One microgram of RNA was reverse-transcribed using dT15-oligonucleotide and Moloney murine leukemia virus reverse transcriptase. To inhibit NADPH oxidase activity, cells were preincubated with DPI. Real-time quantitative RT-PCR was done using a TagMan System (Applied Biosystems, Foster City, CA), using the following primers: TGF-β1; forward 5′-TCTCTCCGACCTGCCACAGA-3′, reverse 5′-GATCGCGCCCATCTAGGTT-3′, procollagen α(1) I; forward 5′-AAAGCAGAAACATCGGATTTGG-3′, reverse 5′-CGTGTCATCCCTTGTGCCGCA-3′. The results were expressed as a ratio of product copies/mL to copies/mL of the housekeeping gene (GAPDH) from the same RNA sample and PCR run.

RNA Interference.

The target sequence to rat TGF-β1 small interfering RNA (siRNA) was AACAACGCAATCTATGACAAA (Qiagen, Valencia, CA), which corresponds to nucleotides 761-780 of rat TGF-β1 mRNA (NM #021578). For negative control, labeled (Alexa Fluor 488) or unlabeled siRNA was used (Qiagen, Valencia, CA). Primary rat stellate cells were cultured and when they reached log phase growth approximately 60% confluence, were treated with 50 nmol/L siRNA in RiboJuice (Novagen, Madison, WI), according to the manufacturer's instructions. Random siRNA in the same concentration was used as control.

Transmission Electron Microscopy.

Specimens (human liver biopsies, BDL rat liver, and micropig liver) were pre-fixed in Karnovsky's fixative (2.5% glutaraldehyde and 2.0% paraformaldehyde) in 0.08 mol/L phosphate buffer (pH 7.3).22 The tissues were secondarily fixed in 1% freshly prepared osmium tetroxide and 1.5% potassium ferrocyanide in double distilled water for 1 hour. After being washed with cold double distilled water 3 times, the tissue blocks were dehydrated through an ascending concentration of ethanol (30%, 50%, 70%, 95%, and 100%), and three changes of 100% ethanol, embedded in a 100% Epon-araldite resin, and polymerized at 70°C. The embedded blocks were sectioned using a diamond knife on a Leica, Ultracut UCT (Leica Microsystems, Deerfield, IL). Sections were placed on a coated grid, and stained with uranyl acetate and lead citrate before being examined in the Philips 120 BioTwin Electron Microscope at 80KV (Fei Company, Hillsboro, OR) equipped with a Gatan/MegaScan, model 794/20 digital camera (Gatan Inc., Pleasanton, CA).

Statistical Analysis.

All data represent at least three experiments and are expressed as the mean ± SEM. Differences between groups were compared using ANOVA for repeated measures and the post-hoc Bonferroni test to correct for multiple comparisons. A P value of less than .05 was considered statistically significant.


Phagocytosis of Apoptotic Bodies Induces NADPH Oxidase and Release of Intra- and Extracellular Superoxide Anions.

To assess the impact of phagocytosis of AB on superoxide production, AB were generated by exposing HepG2 cells to UV radiation (100 mJ/cm2), and incubated with LX-1 cells, an immortalized human HSC line, either in the presence or absence of diphenylene iodonium (DPI, 10 μmol/L for 30 minutes). DPI is a flavoprotein inhibitor that attacks one of the components of the NADPH oxidase complex, and inhibits flavocytochrome b558, and thus inhibits NADPH oxidase activity. Based on the cytochrome c reduction assay, phagocytosis of AB significantly increased superoxide anion production and release in both the extracellular and intracellular milieus (Fig. 1). The extracellular release of superoxide anion (Fig. 1A) increased by 17.4-fold (1 hour), 22.2-fold (6 hours), and 15.1-fold (24 hours). Preincubation with DPI inhibited the release of superoxide. Macrophages were used as a positive control for phagocytosis-induced superoxide production, whereas LX-1 cells without AB exposure served as a negative control to establish the level of baseline superoxide anion production in LX-1 cells. In parallel, intracellular production of superoxide was also increased significantly after engulfment of AB, by 11.4-fold (1 hour), 27.4-fold (6 hours), and 22.6-fold (24 hours) compared with the negative control. Again, DPI prevented superoxide production, the cells producing only a baseline amount of superoxide. These biochemical studies verified the hypothesis that engulfment of AB by HSC indeed induces a significant production of superoxide anion, with release into both the intracellular and extracellular milieus.

Figure 1.

Phagocytosis of AB induces significant increase in superoxide production and release into the (A) extra, and (B) intracellular milieu. This is measured at 1 hour, 6 hours, and 24 hours after the addition of AB. Preincubation of the cells with the NADPH oxidase inhibitor DPI (10 μmol/L for 30 minutes) prevents production of superoxide. Macrophages were used as a positive control. LX-1 cells without AB were the negative control. Four different experiments *P < .05. AB, apoptotic bodies; DPI, diphenylene iodonium.

Confocal Microscopy for Detection of Oxidative Radicals in HSC Phagocytosing AB.

We next used a morphological approach to co-localize labeled AB and oxidative radicals within LX-1 cells. After cells engulfed TAMRA-labeled AB, they were incubated with the redox-sensitive dye, DCFDA, and the images were analyzed by confocal microscopy (Fig. 2). An increase in DCF signal was noted 6 hours after exposure to AB. Fig 2A depicts the co-localization of the DCF signal (green) and intracellular AB (red). Preincubation with DPI reduced the production of oxidative radicals, corresponding to our biochemical data (Fig. 2B). LX-1 cells without exposure to AB exhibited only low levels of DCF signal (Fig. 2C). Macrophages were used as a positive control, and after exposure to AB, intense fluorescent signal was noted in these cells (Fig. 2D). These data therefore corroborate our previous biochemical results, showing that engulfment of AB in HSC induces production of oxidative radicals. DCFDA, however, is not specific for superoxide anion detection and NADPH oxidase activation. Therefore, we complemented these findings with a morphological method demonstrating that phagocytosis of AB induces NADPH oxidase.

Figure 2.

Phagocytosis of AB by HSC increases production of oxidative radicals. Six hours after exposure to AB, an increase in DCFDA signal noted in HSC as assessed by confocal microscopy. (A) Represents the DCFDA signal (green) and intracellular AB (red). (B) Preincubation with 10 μmol/L DPI for 30 minutes decreased the production of oxidative radicals. (C) Control HSC (without exposure to AB) with baseline DCFDA signal. (D) Macrophages used as positive control. Bar = 10 μm. AB, apoptotic bodies; HSC, hepatic stellate cells; DCFDA, 2′,7′-dichlorofluorescein diacetate; DPI, diphenylene iodonium.

p47phox Translocation in HSC Phagocytosing AB.

To specifically document activation of NADPH oxidase, TAMRA-labeled AB were added to HSC as described, then immunohistochemistry was performed to detect the intracellular location of the p47phox subunit of NADPH oxidase. On activation of the enzyme, the p47phox subunit migrates to the cell membrane (Supplementary Fig. 1a; Supplementary material can be found on the HEPATOLOGY website: Next, HSC were preincubated with ML-7, an inhibitor of the myosin light chain kinase, known to prevent migration of the cytosolic subunits to the membrane, and therefore NADPH oxidase activation. Indeed, this inhibited translocation of the p47phox to the membrane, and resulted in only a cytosolic signal, similar to the control cells (Supplementary Fig. 1b). In control HSC, without exposure to AB, p47phox is cytosolic (Supplementary Fig. 1c). These experiments therefore confirmed that phagocytosis of AB by HSC results in NADPH oxidase activation.

Phagocytosis of AB Induces Procollagen α1 (I) and TGF-β1 Expression.

Expression of procollagen α1 (I) and TGF-β1 mRNAs was quantified after 48 hours incubation with AB. To inhibit NADPH oxidase, cells were preincubated with DPI before addition of the AB. Expression of procollagen α1 (I) increased significantly in cells phagocytosing AB; this was decreased by inhibiting NADPH oxidase, suggesting that upregulation of the procollagen α1 (I) in this case was NADPH-oxidase and superoxide-dependent (Fig. 3A). Conversely, although phagocytosis of AB induced upregulation of TGF-β1, this was not prevented by inhibiting NADPH oxidase, indicating that TGF-β1 signals by a different mechanism after phagocytosis, which does not involve NADPH oxidase activation (Fig. 3B).

Figure 3.

Real-time PCR of procollagen α1(I) (A) and TGF-β1 (B) expression in LX-1 cells. (A) Phagocytosis of AB induces upregulation of procollagen α1(I). This is decreased by the preincubation of LX-1 cells with DPI (10 μmol/L for 30 minutes). (B) Phagocytosis of AB induces upregulation of TGF-β1. This is not inhibited by the preincubation of LX-1 cells with DPI, suggesting that upregulation of TGF-β1 occurs by an NADPH oxidase–independent pathway. The expression of procollagen α 1(I) and TGF-β1 is normalized to the expression of GAPDH. LX-1 cells without exposure to AB used as controls. Three different experiments; *P < .05. PCR, polymerase chain reaction; TGF-β1, transforming growth factor beta1; DPI, diphenylene iodonium; AB, apoptotic bodies.

Inhibition of TGF-β1 Expression Does Not Inhibit Superoxide Production After Phagocytosis.

RNA interference using siRNA was used to inhibit expression of TGF-β1 in primary rat stellate cells. Incubation with 50 nmol/L siRNA decreased TGF-β1 expression by 62% in cells exposed to apoptotic bodies (Fig. 4A). Procollagen α1 (I) expression decreased by 30% after incubation with TGF-β1 siRNA (Fig. 4B). Superoxide production induced by phagocytosis decreased only by 13% after incubation with the TGF-β1 siRNA (Fig. 4C). The data from these experiments support our previous findings and are suggestive of two different signaling pathways induced by phagocytosis of apoptotic bodies: an NADPH oxidase-dependent pathway, and an NADPH-oxidase independent pathway involving TGF-β1 activation.

Figure 4.

Inhibition of TGF-β1 expression in primary rat HSC by siRNA. Primary rat stellate cells were incubated with 50 nmol/L siRNA to TGF-β1 and then exposed to AB. As control, transfection with random siRNA was performed. (A) Expression of TGF-β1 was inhibited by 62%. (B) Expression of procollagen α 1(I) was inhibited by 30% by incubation with TGF-β1 siRNA. (C) Production of superoxide showed no significant change after incubation with TGF-β1 siRNA. *P < .05, average of four experiments for (A), three experiments for (B), and four experiments for (C). TGF-β1, transforming growth factor beta1; HSC, hepatic stellate cells; siRNA, small interfering RNA.

Phagocytosis of AB in Different Liver Fibrosis Models.

To ascertain whether phagocytosis of AB by HSC occurs in vivo as well, we studied EM sections of livers of rats 7 days after BDL, micropigs fed with FDE diet for 14 weeks23 and human liver biopsy samples with recurrent post-transplantation hepatitis C. The biopsy samples were from three different patients with stage 1 fibrosis. Mechanistically distinct models (cholestatic, alcohol-induced, and viral) were employed in different species to establish that these findings were generic to all forms of fibrotic stage liver disease. Activated HSC lose their fat droplets and have well-developed rough-surfaced endoplasmic reticulum and large Golgi apparatus. An increased amount of microfilaments is seen underneath the plasma membrane, a feature of myofibroblasts or activated HSC.2, 24–26 Activated HSC localize in close proximity to collagen fibers. In vivo, engulfed AB become phagosomes, and at a later stage, phagolysosomes that are readily identified as double membrane-bound intracytoplasmic organelles containing remnants of cell organelles, and chromatin derived from apoptotic cells. Phagocytosed necrotic debris morphologically would appear different because the membrane integrity is lost during necrosis: therefore, the uptake does not occur as discrete particles, and a double membrane would not be found around them.27 In all three models many HSC were seen with phagosomes or phagolysosomes, and these cells displayed condensed bundles of microfilaments, a typical attribute of active HSC (Figs. 5-7). Moreover, the cells did not have fat droplets, but rather had well-developed rough-surfaced endoplasmic reticulum and large Golgi apparatus, and they also appeared to be adjacent to collagen bundles. Specimens from three rats after BDL, two micropigs, and from the three patients with post-transplantation hepatitis C were studied. Reviewing three different sections (each four grids) for every specimen, 33.1% of HSC in human livers with HCV, 30.7% of HSC in BDL rats, and 25.8% of HSC in micropigs fed with FDE diet contained intracellular AB. In contrast, HSC from control micropig liver exhibited the characteristic resting phenotype with lipid droplets, no increased intracellular microfilaments, and intracellular AB rarely were seen (data not shown).6

Figure 5.

HSC in rat liver after BDL. (A) Shows HSC with phagocytosed AB (box). (B) Enlarged image of the AB. Prominent rough ER is seen throughout the cell, characteristic of HSC. (C) Further enlargement of the same HSC shows microfilament bundles (arrow). HSC, hepatic stellate cells; BDL, bile duct ligation; AB, apoptotic bodies; ER, endoplasmic reticulum.

Figure 6.

HSC in pig liver specimen fed with folate-deficient/ethanol diet. (A) Several HSC, collagen bundles (arrowheads), one HSC with phagocytosed AB (arrow). (B) Enlarged image of the AB. (C) Further enlargement of the same HSC reveals prominent microfilament bundles (arrow). HSC, hepatic stellate cells; AB, apoptotic bodies.

Figure 7.

HSC in a liver biopsy specimen from a patient with post-transplantation hepatitis C. (A) HSC contains several intracellular AB (box), as well as prominent microfilaments (arrow). (B) Enlarged image of the AB. It is surrounded by double-layered membrane (arrow) (C) Enlarged image of the microfilaments. HSC, hepatic stellate cells; AB, apoptotic bodies.

Co-localization of Apoptotic Bodies and Activated Stellate Cells in the Micropig Liver.

To address whether the cells with intracellular apoptotic bodies are really stellate cells, we performed double labeling experiments. Micropig liver sections were stained for alpha-smooth muscle actin and E-cadherin. E-cadherin has recently been recognized as a marker for intracellular apoptotic bodies.28 E-cadherin is a membrane protein on epithelial cells, and it is retained on apoptotic bodies derived from these cells.28 Figure 8 shows several stellate cells (arrows) in the micropig liver with intracellular, E-cadherin–positive apoptotic bodies.

Figure 8.

Co-localization of AB and HSC in micropigs fed with FDE diet. E-cadherin–positive AB (red, arrows) in HSC (ASMA, green) in the micropig liver. (A) phase image, bar = 10 μm. (B) Higher magnification of the same field; bar = 20 μm. (C) and (D) Higher-magnification images of different fields (bars = 20 μm and 10 μm, respectively). HSC, hepatic stellate cells; AB, apoptotic bodies.


Our data strongly support a model in which hepatocyte AB are engulfed by HSC, resulting in TGF-β1 and procollagen α1 (I) upregulation and consequently fibrogenesis. Furthermore, NADPH oxidase activation and resulting superoxide production seem to be important components linking phagocytosis to procollagen α1 (I) upregulation. Indeed, p47phox−/− mice do not develop alcoholic liver damage29 and do not develop significant fibrosis after bile duct ligation.30

We found that NADPH oxidase plays a critical role in signaling between phagocytosis and fibrogenesis in the liver. NADPH oxidase activation appears to induce both intracellular (into the phagolysosomes) and extracellular superoxide release. Release of ROS may affect neighboring cells in several ways; for example, it can induce neutrophil extravasation and activation, thereby further exacerbating the inflammatory response.31 ROS can directly damage cells by causing membrane damage through lipid peroxidation32 and mitochondrial dysfunction,33 as well as DNA damage either by direct mutation by DNA strand breaks34 or by impairing the function of DNA repair enzymes.35, 36 These effects may lead to further cell damage, culminating in apoptosis of neighboring hepatocytes,37 thereby reinforcing the cycle of apoptosis, phagocytosis, and fibrogenesis. Our results also indicate that the generation of superoxide induced by the phagocytic process has a direct effect on fibrogenesis. Previously, it was demonstrated that superoxide directly affects procollagen α1(I) expression by a C/EBPB-dependent pathway.38 Also, superoxide production induced by the CYP2EI system was shown to result in collagen 1 upregulation at the transcriptional level.39 Taken together, our data provide a novel mechanism connecting NADPH oxidase activation, production of superoxide, and upregulation of procollagen α1 (I) expression.

The clearance of AB is thought to be an evolutionarily-conserved process to limit inflammatory responses secondary to tissue injury.8, 40 One of the main mediators of the anti-inflammatory effect is TGF-β1, produced by phagocytic cells after ingestion of AB in several systems8, 41 as well as in ours. It is also a potent profibrogenic cytokine in the liver.42 Therefore, excessive apoptosis during ongoing tissue injury of various causes could trigger the fibrogenic response through enhanced TGF-β1 production. This hypothesis may explain recent observations that the detection of caspase activation in sera of patients with hepatitis C as a marker of increased apoptotic activity appears to be associated with a more aggressive fibrogenic process.43 We also observed TGF-β1 upregulation in our culture model of phagocytosis. In contrast to collagen, however, this upregulation does not appear to be dependent on NADPH oxidase activation and superoxide production; yet it appears to be a direct consequence of the phagocytic process itself. Phagocytosis-induced TGF-β1 upregulation may reflect a different signaling pathway. This was confirmed by using siRNA to TGF-β1; inhibiting its expression did not affect phagocytosis-induced superoxide production (Fig. 4). This is in accordance with the recent observations of Svegliati-Baroni et al.,44 showing neutralizing antibodies to TGF-β1 failed to inhibit acetaldehyde-dependent collagen gene transcription.44 Our data show a major effect on procollagen α1 (I) expression after phagocytosis by both NADPH oxidase–dependent and independent pathways.

Although there are no firm data in this area, it is possible that integrins may be involved in the signal transduction during phagocytosis leading to TGF-β1 upregulation. In macrophages after phagocytosis, TGF-β1 induces cross-talk between ERK and p38 MAPK and this mediates selective suppression of pro-inflammatory cytokines.45 Therefore, via these signaling pathways, TGF-β1 may also mediate a fibrogenic stimulus, independent of NADPH oxidase. This is consistent with the findings that inhibition of the p38 MAPK activity reduced type I collagen mRNA levels in primary rodent HSC.46, 47

Our findings demonstrate the presence of AB within HSC in vivo in three different models of fibrosis, confirming that phagocytic activity is indeed a physiologically relevant characteristic of these cells. The phagosomes indeed represent AB, because they are discrete particles surrounded by a double membrane. By immunohistochemistry these intracellular particles are E-cadherin positive, further supporting our data that these are AB. Phagocytosis of necrotic particles could also occur in the liver; however, it is a different phenomenon. Their recognition, tethering, and phagocytosis are described as distinctively different processes compared with the phagocytic process of apoptotic bodies,27, 37 and necrotic particles are not surrounded by a double membrane. HSC are known to transdifferentiate from a resting fat-storing cell into a migratory, contractile myofibroblast on activation. Phagocytic activity is yet another function that these cells can perform, proving their impressive plasticity.

How important is the phagocytosis by HSC in the process of liver fibrogenesis? HSC are located in close proximity to hepatocytes; therefore, they may be more apt to phagocytose AB from hepatocytes than the more distant Kupffer cells. Furthermore, phagocytosis by HSC seems to directly upregulate profibrogenic genes. Our in vivo data, detecting phagocytosing HSC in livers, provide evidence about the relevance of this finding in the disease process. Thus, phagocytosis by HSC is likely to be a common, universal pathway leading to fibrosis independent of the etiology of the liver disease.

Therefore, our studies indicate that phagocytosis of AB by HSC is an important process in fibrogenesis, linking chronic liver injury and apoptosis of hepatocytes to fibrogenesis. We must develop a better understanding of how the signaling pathways induced by phagocytosis result in fibrogenesis. Preventing phagocytosis-induced signaling events could be a promising approach to develop successful therapy for liver fibrosis.


The authors thank Grete Adamson for providing assistance with the EM studies and Dr. Andreea Catana for expert assistance.