|Home | About | Journals | Submit | Contact Us | Français|
Joy X. Jiang: performing experiments and writing paper
Senthil Venugopal: performing experiments
Nobuko Serizawa: performing experiments
Xiangling Chen: performing experiments
Fiona Scott: performing experiments
Yong Li: performing experiments
Roger Adamson: technical support
Sridevi Devaraj: technical support
Vijay Shah: collaboration, technical support
M. Eric Gershwin: critical revision of the manuscript for intellectual content
Scott L Friedman: critical revision of the manuscript for intellectual content
Natalie J Török: concept and design, analysis of the data, funding
Hepatocyte apoptosis and activation of hepatic stellate cells (HSC) are critical events in fibrogenesis. We previously demonstrated that phagocytosis of apoptotic hepatocytes by HSC is profibrogenic. Based on this, as well as the observation that NADPH oxidase induction is central to fibrogenesis, our aim was to study the phagocytic NADPH oxidase, NOX2.
An in vivo phagocytosis model was developed by injecting wild type (wt) or NOX2-/- mice with lentiviral-GFP containing a hepatocyte-specific promoter, and ad-TRAIL. Fibrosis was evaluated in bile duct ligated (BDL) wt and NOX2-/- mice with or without gadolinium treatment. NOX2 expression was studied in human liver samples and in HSC isolated from fibrotic livers. The fibrogenic activity of NOX2 was assessed by collagen reporter assays.
In the phagocytosis model engulfment of GFP-labeled apoptotic bodies was seen, and the expression of α-SMA and collagen I increased significantly in the wt, but not in the NOX2-/- mice. Inhibiting apoptosis decreased the profibrogenic response. NOX2-/- animals exhibited significantly less fibrosis following BDL. Inactivating macrophages in wt BDL mice did not lower collagen production to the level observed in NOX2-/- mice suggesting that NOX2-expressing HSC are important in fibrogenesis. NOX2 was upregulated in HSC from fibrotic livers, and phagocytosis-induced NOX2 expression and activity were demonstrated. Based on reporter assays, NOX2-mediated ROS directly induced collagen promoter activity in HSC.
Apoptosis and phagocytosis of hepatocytes directly induce HSC activation and initiation of fibrosis. NOX2, the phagocytic NADPH oxidase plays a key role in this process and in liver fibrogenesis in vivo.
Liver fibrosis is the end result of chronic liver injury of various etiologies1. Apoptosis of hepatocytes is a common feature of this process independent of the type of liver injury2. The apoptotic cells are usually efficiently eliminated by the professional phagocytes by efferocytosis however, when the phagocytic system is overwhelmed, non professional phagocytes start to increasingly play a role in phagocytosis3. We have recently identified hepatic stellate cells (HSC) performing phagocytic function during chronic liver injury4. HSC are central to liver fibrosis; upon activation they transdifferentiate into myofibroblasts, produce extracellular matrix (ECM) components and profibrogenic cytokines. Phagocytosis of hepatocyte-derived apoptotic bodies (AB) by HSC is profibrogenic as it induces collagen α1(I) and TGF-β1 upregulation, and NADPH oxidase (NOX)-dependent superoxide production4. In addition, phagocytosis and NADPH oxidase-mediated signaling events induce myofibroblast survival5. Thus, phagocytosis and resulting HSC activation could be amongst the first, initiating events in liver fibrogenesis linking hepatocyte apoptosis to HSC activation and production of ECM. HSC express the p22phox, p47phox, p67phox and gp91phox (NOX2), subunits of the NADPH oxidase enzyme complex, as well as NOX16. During activation of the phagocytic NOX, the cytosolic subunits (p47phox, p67hox) migrate to the membrane, where they associate with the subunits gp91phox and p22phox to assemble the active oxidase. Gp91phox is the rate-limiting catalytic component, and its upregulated transcription was shown to increase respiratory burst capacity during inflammatory responses7. Rac GTPases are also important regulatory elements of NOXs8. Upon activation, Rac-GTP translocates to the membrane where it interacts with p67phox. Rac1 is ubiquitously expressed while Rac2 is only expressed in myeloid cells9. Rac1 was shown to activate NOX2 in heterologous systems resulting in reactive oxidative species (ROS production)8. Expression of the constitutive active Rac1 induced accelerated liver fibrosis in mice10, underlying the importance of Rac1 in fibrogenesis. On the other hand, Rac1 activation is also linked to the induction of phagocytosis11. Several profibrogenic stimuli can activate NOXs in HSC, such as PDGF12, angiotensin II13, and leptin14. However, in HSC the mechanism of activation, function and downstream signaling events elicited by the different NOXs have not yet been clearly elucidated.
In this study we have shown that apoptosis of hepatocytes is directly linked to stellate cell activation in vivo, via the activation of NOX2 (gp91phox) in HSC. This represents an essential early event in liver fibrogenesis with a direct effect on ROS-mediated collagen upregulation in HSC. NOX2 thus is a central enzyme in the induction of early fibrogenic changes.
The liver biopsy samples were obtained from the UC Davis Cancer Center Biorepository funded by the NCI.
In vivo phagocytosis model: wt (C56BL6) or NOX2-/- mice (same background, Jackson Lab, Bar Harbor, ME) were injected with lentiviral (LV)-GFP with the hepatocyte-specific α1 antitrypsin (α1-AT) promoter (3×107pfu/g) (gift from Dr. Zern, UC Davis) into the portal vein, then 7 days later Ad-TRAIL (2×107pfu/g) (gift from Dr. Gores, Mayo Clinic), into the tail vein of the same mice. A separate group of mice were treated by i.p. injection of the pancaspase inhibitor Q-VD-OPH (10 mg/kg, MP Biomedicals, Solon, OH; twice on the day of Adeno-TRAIL injection and the day after), and controls were injected with the vehicle, DMSO. As for control injections, mice were injected either with α1-AT-LV-GFP (3×107 pfu/g), or Ad-TRAIL (2×107pfu/g), or Ad-GFP (2×107pfu/g, Vector Biolabs, Philadelphia PA), only. The animals were sacrificed 3 days after the Ad-TRAIL injection.
BDL was performed on wt and NOX2-/- mice, as described4. Sham operation was conducted in parallel. Gadolinium chloride (GdCl3, 10 mg/kg in saline, Sigma-Aldrich) was injected i.p. every other day throughout the experiment. The mice were sacrificed 3 or 6 weeks after surgery. To study NOX2 expression HSC were isolated from BDL or sham-operated Sprague Dawley rats (Charles River Laboratories Inc., Wilmington, MA) as described below.
All data represented at least three experiments and expressed as the mean ± SED. Differences between groups were compared using one-way Analysis of Variance (ANOVA) associated with the Dunnett's test. Statistical significance was assumed when p<0.05.
For details regarding histology, immunohistochemistry, TUNEL assay, hydroxy-proline assay, serum biochemical measurements, preparation of apoptotic bodies and phagocytosis experiments, virus preparation, siRNA experiments, real time PCR, lucigenin assay, reporter assays assessing collagen promoter activity, and Rac1 pull down assay, western blotting please see Supplementary Materials and Methods.
We performed immunohistochemistry and confocal microscopy to study NOX2 expression in liver specimens of patients with HCV or PBC. NOX2 was expressed in α-SMA positive HSC (Figure 1A) in both HCV and PBC. To recapitulate this in an animal model of liver fibrosis, BDL was performed in rats, and HSC were isolated (the purity of the HSC was confirmed by α-SMA staining). NOX2 mRNA (12.7±1.2 fold, *p<0.05) and protein expression (4.22±0.004 fold, p<0.05, densitometry, not shown) were significantly increased in primary HSC from BDL rats while low levels of expression were seen in HSC from sham-operated animals (Figure 1B). As the signals inducing the phagocytic NOX2 expression in activated HSC are not known, we postulated that engulfment of apoptotic bodies (AB) could be a trigger. To test this, primary HSC were exposed to AB and NOX2 expression was tested by real-time PCR. We found that NOX2 expression was significantly induced (4.25±0.98 fold, **p≤0.0001) in HSC following phagocytosis (Figure 1C). Taken together with our earlier data, upregulation of NOX2 following phagocytosis may translate into increased collagen I production by HSC.
Primary rat HSC were transfected with a NOX2 siRNA or scrambled siRNA, then exposed to AB. Superoxide production was assessed by the lucigenin assay (Figure 2A). In the scrambled siRNA-transfected cells phagocytosis induced superoxide production (1.75±0.24 fold, *p<0.05), and this was inhibited in the NOX2 siRNA-transfected cells. Next, we assessed if collagen upregulation was NOX2-mediated following phagocytosis by real-time PCR using scrambled or NOX2 siRNA-transfected HSC. Collagen IA1 expression was upregulated in the scrambled siRNA-transfected HSC following AB exposure (**p≤0.01), and this decreased significantly in the NOX2 siRNA-transfected primary HSC (by 88%±4, ##p≤0.0001), suggesting that this is a central enzyme in phagocytosis-induced fibrogenic responses (Figure 2B).
Collagen consists of two α1 chains and one α2 chain. Both COLIA1 and COLIA2 genes are highly sensitive to ROS15. The promoters contain a H2O2-responsive area therefore we studied whether the NOX2-mediated superoxide and peroxide production could directly result in collagen promoter activity. Primary wild type or NOX2-/- HSC were transfected by the constructs containing the truncated promoter Col1A2 P1-Luc (-378/+58), or with a construct where the peroxide-sensitive area was intact [Col1A2 P1-Luc (-2900/+58)], or empty vector. The cells were then exposed to AB in the presence or absence of the reducing agent GSH or catalase. Engulfment of AB by HSC resulted in a significant induction of the COLIA2 promoter activity in the Col1A2 P1-Luc (-2900/+58)-transfected cells (35.7±6.2 fold, **p≤0.01), compared to control cells. This was abrogated in the Col1A2 P1-Luc (-378/+58)-transfected cells (1.95±0.21 fold, **p≤0.01), or decreased after exposure to catalase (5.66±0.4 fold, **p≤0.01), indicating that the promoter activity resulted from peroxide produced following phagocytosis (Figure 2C). In NOX2-/- cells the luciferase activity was significantly blunted following phagocytosis (1.71±0.21 fold, *p<0.05). NOX2 hence is a candidate enzyme regulating superoxide and peroxide-mediated induction of collagen expression during fibrogenesis.
In our previous studies we found that phagocytosis of AB induced Rac1 activation, an important regulatory element of NOX2 in non-hemopoietic cells; and constitutive active Rac1 also augmented the phagocytic activity of HSC16. Rac recruitment to the enzyme complex is known to be significantly decreased in CGD neutrophils (with the mutation of the gp91phox) 17. Thus it is plausible that intact NOX2 is required for Rac1 recruitment and activation in HSC for phagocytosis to occur. To test this, first, the phagocytic rate in wt and NOX2-/- HSC was studied, and we found that NOX2-/- HSC engulfed significantly less AB (#p<0.001, Figure 3A). To assess whether this decline in the phagocytic activity was actually due to decreased GTP-ase activity and/or decreased recruitment of Rac1 to the enzyme complex in the membrane, we performed Rac1 pull-down assays in wt and NOX2-/- HSC following exposure to AB and tested the membrane fractions for GTP-Rac1. In the NOX2-/- HSC the amount of active Rac1 in the membrane fraction was decreased (Figure 3B). While the antibody against active Rac1 labeled the phagosomal membrane in wt cells, the phagosomal labeling was not seen in NOX2-/- cells. Tethering of TAMRA-labeled AB in NOX2-/- cells did occur, but the engulfment did not take place or was often incomplete. Intact NOX2 is thus required for Rac1 recruitment, activation and for phagocytosis. As phagocytosis could be an important early event in hepatocyte apoptosis-induced fibrogenic activity, we next will test the validity of these in vitro data by a novel model of in vivo phagocytosis.
In this study we induced selective apoptosis of GFP positive hepatocytes to track the fate of their AB. This model was based on the notion that TRAIL is not apoptotic on normal hepatocytes, but it induces cell death of virus-infected hepatocytes18-19. To track apoptotic hepatocytes we used a lentiviral approach where GFP only expressed by hepatocytes because of the expression of the hepatocyte-specific promoter α1-AT. The animals were first injected with the α1-AT-LV-GFP via the portal vein then 7 days later Ad-TRAIL was injected into the tail vein of the same mice. TRAIL-mediated apoptosis of GFP positive, virus-infected hepatocytes was then detected. As control, mice were injected only with α1-AT-LV-GFP, or only with Ad-TRAIL, or Ad-GFP. A separate group of mice were injected with the pancaspase inhibitor Q-VD-OPH prior to the viral injections. The liver tissues from only Ad-TRAIL, or LV-injected mice showed no significant injury or infiltration by inflammatory cells, and the ALT values remained normal (Figure 4b, c). In the LV plus Ad-TRAIL-injected animals the liver showed mild to moderate hepatocyte injury with associated regenerative activity, increased ALT values, and no significant infiltration by inflammatory cells (Figure 4d). In the pancaspase inhibitor-treated LV plus Ad-TRAIL-injected group, the liver histology was normal (Figure 4e), and the ALT decreased significantly (p≤0.05). To validate our model, and demonstrate apoptosis of hepatocytes, TUNEL assays were done on all samples. In the LV plus Ad-TRAIL infected livers hepatocyte apoptosis was increased (Figure 4d’) compared to only Ad-TRAIL (Figure 4b’), or only LV (Figure 4c’) injected animals. In the Q-VD-OPH treated group the number of apoptotic cells has decreased (Figure 4e’). Studying liver histology in each experimental condition, we have not seen infiltration by inflammatory cells. To confirm this, immunohistochemistry for CD11b (Mac-1) was done and we found similar numbers of CD11b positive cells in each experiment (Supplementary Figure 1A). In addition, real-time PCR showed comparable levels of TNF-α expression in each condition.
To analyze phagocytosis, immunohistochemistry and confocal microscopy were performed to visualize GFP-labeled hepatocytes and activated HSC (α-SMA, red). There was an increase in the number of activated HSC in the liver of mice injected with α1-AT-LV-GFP plus Ad-TRAIL (Figure 5d, e), while in the only LV (Figure 5c), or only Ad-TRAIL (Figure 5b) injected mice no increase was detected similar to control (Figure 5a). At higher magnification in the α-SMA-positive HSC, GFP-labeled AB could be seen (arrow), indicating phagocytosis of hepatocyte-derived AB (Figure 5f). Apoptosis and phagocytosis mediated fibrogenic changes were also confirmed in a different model where wt and galectin 3-/- mice were treated as above, and a decrease in procollagen α 1(I) and TGF β expression were seen in the galectin 3-/- mice (manuscript in preparation). Galectin 3 is necessary for phagocytosis by facilitating of the tethering of AB20.
Apoptosis of hepatocytes thus directly induced HSC activation in vivo as no fibrogenic agents or methods were used in this model. This directly supports the hypothesis that apoptosis of hepatocytes is profibrogenic.
Based on our in vitro data, in NOX2-/- HSC, the upregulation of collagen was blunted. To assess this in our in vivo model both wt and NOX2-/- mice were injected with Ad-TRAIL, or LV or LV plus Ad-TRAIL, with or without the pancaspase inhibitor, as above. Immunohistochemistry showed less α-SMA positive HSC in NOX2-/- mice after injection of LV plus Ad-TRAIL, suggesting that in these animals HSC activation was blunted (images for wt and NOX2-/- injected with LV plus Ad-TRAIL are shown, with or without the pancaspase inhibitor, Figure 6A). To confirm these data, real-time PCR was performed on the livers of virus-injected wt and NOX2-/- mice using α-SMA, collagen IA1 and TGF-β1 specific primers. The expression of α-SMA (31.3±6.16-fold, *p<0.05), collagen IA1 (6.03±0.7-fold, **p≤0.01), and TGF-β1 (17.03±2.6-fold, *p<0.05) have significantly increased in LV plus Ad-TRAIL injected wt animals (Figure 6B), while no increase was detected in NOX2-/- animals. In addition, treatment with the caspase inhibitor decreased the upregulation of the fibrogenic markers significantly (*p<0.05). Taken together, these data indicate that NOX2 is essential in the early upregulation of pro-fibrogenic genes following the apoptosis of hepatocytes.
To further study the in vivo relevance of the above findings, BDL was performed in wt and NOX2-/- mice. We chose BDL as fibrosis-inducing method, as both carbon tetrachloride (CCl4) and thioacetamide (TAA)-induced liver injury cause significant oxidative stress and hepatocyte necrosis which may confound the data. The animals were sacrificed after 3 weeks and in the case of some NOX2-/- mice after 6 weeks following surgery, and the liver specimens were processed for picrosirius staining to assess fibrosis stage (wt animals did not survive 6 weeks after BDL). The effects of BDL were similar in the wt and NOX2-/- livers, showing the same degree of inflammation (grade 2) and bile duct proliferation (data not shown), and apoptosis (TUNEL assay, data not shown). We found that in NOX2-/- animals the fibrosis stage was significantly lower compared to that of wt animals following BDL (Figure 7A b, c, d). The serum was tested for ALT and bilirubin values and we found that both ALT and bilirubin were increased in BDL animals (*p<0.05); (in both wt and NOX2-/-, with the ALT somewhat lower in NOX2-/- mice than in wt albeit with no statistical significance). To elucidate the relative contribution of HSC and Kupffer cells to liver fibrosis in respect to their NOX2 expression, mice undergone BDL were injected with GdCl3, throughout the experiment, to inhibit macrophages. Fibrosis stage was lower in GdCl3-injected wt animals after BDL compared to PBS-injected controls (Figure 7B a, compared to figure 7A, b), consistent with previous data 21. To assess collagen expression, and liver collagen content, real time PCR and OH proline assays were performed in all experimental conditions. Collagen expression (55%±6.8, **p≤0.01) and OH-proline incorporation (63%±9.5, *p<0.05) have significantly decreased in NOX2-/- mice compared to wt animals following BDL. In the GdCl3-injected BDL wt mice the expression of collagen IA1 and incorporation of OH-proline have decreased to a certain extent (24%±2.3,*p<0.05, and 42%±5.6, *p<0.05), compared to PBS-injected BDL animals. However, this decline in fibrogenic activity following macrophage inhibition in wt animals (white bars) did not reach the low level observed in NOX2-/- animals (black bars), suggesting that NOX2 expressing HSC may play a major role in liver fibrosis.
In this study we have shown that 1) phagocytosis of apoptotic hepatocytes is directly profibrogenic in vivo, 2) the profibrogenic effect is mediated by NOX2 and 3) NOX2-/- animals have reduced fibrosis. Phagocytosis of apoptotic cells is essential in maintaining tissue homeostasis. According to current concepts, engulfment of AB in physiological circumstances is anti-inflammatory22. In pathological situations however, such as chronic liver disease hallmarked by ongoing apoptosis of hepatocytes, non professional phagocytes such as HSC start engulfing apoptotic cells. The notion that cells of non-myeloid lineage can phagocytose is not novel: it has been described that epithelial23 or mesenchymal cells24 can engulf AB. Here we have shown that HSC can phagocytose apoptotic hepatocytes and directly induce fibrogenic responses by an ROS-mediated collagen production. Pivotal to this process was the activation of the phagocytic NADPH oxidase, NOX2. An important correlation to the role of NOX2 is the liver disease of patients with CGD with mutation of the components of NOX2, most commonly gp91phox. In these patients the liver is affected by recurrent infections, vascular abnormalities, and eventually non-cirrhotic portal hypertension may develop25. The fact that these patients do not develop liver fibrosis in the face of chronic inflammation is intriguing and pointing to the important role of NOX2 in liver fibrogenesis. Previously, by the elegant studies of Bataller et al, it was shown that NADPH oxidase activation was indeed necessary for angiotensin II-induced liver fibrogenesis13. In those studies p47phox-/- mice were used, a subunit known to be an organizer for both NOX1 and NOX226. Deficiency in NOX2 was shown to increase hepatocellular injury (mainly necrosis) in the CCl4-induced model of fibrosis, upregulation of collagen expression but interestingly, decrease in fibrosis27. In that model increased matrix metalloproteinase 2 (MMP2) and 9 expression in NOX2-/- animals were thought to lead to a decrease in fibrosis. CCl4 is known to induce ROS-mediated liver injury (independent of NOXs) with lipid peroxidation, and consequent necrosis. Thus, the mechanism of liver injury is distinct compared to our model, where ROS production is rather a consequence than a cause of ongoing apoptosis and resulting HSC activation.
ROS are known to play a key role in HSC activation15, 28. H2O2 derived from hepatocytes induced collagen transcription in HSC15 and we have shown that NOX-derived ROS induce survival pathways in HSC contributing to the propagation of activated HSC5. Here we demonstrated that NOX2 activation and peroxide production directly resulted in the induction of the collagen αI promoter in HSC. An important corollary to NOX2 activation is Rac1, an essential subunit of the enzyme and a positive regulator of phagocytosis. Constitutive activation of Rac1 lead to accelerated liver fibrosis emphasizing the role of Rac1 in ROS-mediated liver injury10, and Rac1 was shown to play an important role in the phagocytosis of lymphocytes during fibrogenesis29. In our study we found impaired translocation of Rac1 to the membrane in NOX2-/- HSC, consistent with previous reports30. Thus decrease in the GTP-bound Rac1 at the site of the phagosome in NOX2-/- cells may translate into less effective engulfment. This, together with the decrease in collagen promoter activation in NOX2-/- HSC may translate into a significant reduction in fibrogenic activity. As phagocytosis may represent an early “initiator” event in fibrogenesis, it was essential to recapitulate this in vivo. In our model TRAIL-mediated apoptosis of hepatocytes induced α-SMA and production of collagen IA1 and TGF-β1 in wt mice. While phagocytosis of apoptotic cells is a direct profibrogenic stimulus, we can not exclude other mechanisms of fibrosis in this model. It is possible that increase in cell death also predisposes to damage-associated molecular patterns (DAMP)-mediated liver injury and HSC activation via the toll like receptors (TLR). This was demonstrated earlier by CpG DNA-mediated TLR9 induction on HSC and their subsequent activation 31. In NOX2-/- mice the fibrogenic activity was not seen, confirming the role of NOX2 in early fibrosis. As Kupffer cells express NOX2, their role in early fibrosis had to be addressed. To inhibit Kupffer cells we used GdCl3 throughout the BDL experiment, to avoid repopulation of macrophages. We found that in GdCl3-treated wt animals the fibrosis and collagen production was decreased after 3 weeks, but did not reach the low, almost baseline level of collagen production seen in BDL NOX2-/- mice. This suggests that beside macrophages, HSC via their NOX2 expression and activity are important contributors to early fibrogenic events. Other cells in the liver could potentially also contribute to liver fibrosis via their NOX2 activity, and this could be a focus of future studies. Our main goal here was to study the early, initiating events in fibrosis. At a later, propagation stage of fibrosis however, NOX2 could be also activated by inflammatory mediators or cytokines such as angiotensin II, leptin or PDGF further accelerating the production ECM.
In summary, based on our in vitro and in vivo data, NOX2 is a central enzyme in liver fibrosis. It is especially important in the initiating phase of fibrogenesis when phagocytosis of apoptotic cells is one of the main profibrogenic events. Targeted inhibition of NOX2 activation may prove to be a powerful new strategy to inhibit multiple profibrogenic pathways and halt the progression of the disease.
We would like to thank Dr. Gregory Gores (Mayo College of Medicine) for supplying the Ad-TRAIL construct and to Drs. Mark Zern (UC Davis) and Antonia Follenzi (Albert Einstein School of Medicine) for supplying the lentiviral vector.
Grant Support: This study was supported by the NIH DK069765, DK080715 DK083283 (NJT) and the ALF Post-doctoral Research Awards (JXJ).
Disclosures: Nothing to disclose