Potential conflict of interest: Nothing to report.
The impact of Kupffer cells (KCs) on the hepatic stellate cell (HSC) fibrogenic response was examined in an in vitro coculture model of primary KCs and HSCs. Coculture with KCs induced a more activated phenotype and greater proliferation compared to HSC cultured alone. Similar results were obtained on Matrigel which maintains HSCs quiescent. The effect of KCs on HSC collagen I involved transcriptional regulation, as determined by nuclear in vitro transcription run-on assays, promoter studies, and Northern blot analysis, while stability of the COL1A1 and COL1A2 mRNA were similar. The minimal COL1A1 and COL1A2 promoter regions responsible for the KC effects were localized to the −515 and −378 base pair (bp) regions, respectively. Intracellular and extracellular collagen I protein, H2O2, and IL-6 increased in a time-dependent fashion, especially for HSCs in coculture. Catalase prevented these effects as well as the transactivation of both collagen promoters. The rate of collagen I protein synthesis and intracellular collagen I degradation remained similar but the t1/2 of the secreted collagen I was lower for HSC in coculture. MMP13, a protease that degrades extracellular collagen I, decreased in the cocultures, while TIMP1, a MMP13 inhibitor, increased; and these effects were prevented by catalase, anti-IL-6, and siRNA-IL-6. Cocultured HSC showed elevated phosphorylation of p38 which when inhibited by catalase, anti-IL-6, and siRNA-IL-6 it blocked TIMP1 upregulation and collagen I accumulation. In conclusion, these results unveil a novel dual mechanism mediated by H2O2 and IL-6 by which KCs may modulate the fibrogenic response in HSCs. (HEPATOLOGY 2006;44:1487–1501.)
Efforts to understand liver fibrosis focus primarily on events that lead to activation and proliferation of hepatic stellate cells (HSCs) and the early accumulation of scar, e.g., collagen type I, in hope of identifying therapeutic targets to prevent its establishment, slow its progression, or help its resolution.1 Stimuli initiating HSC activation may derive from injured hepatocytes, Kupffer cells (KCs), and endothelial cells, in addition to rapid changes in ECM composition.1
KCs release a wide array of soluble mediators, including reactive species, cytokines, growth factors, cyclooxygenase and lipoxygenase metabolites, all of which provide physiologically diverse and pivotal paracrine effects on all other liver cells.2, 3 KCs are also central to the liver's homeostatic response to injury, as upon degenerative changes in hepatocytes they immediately respond to the insult and release mediators to orchestrate inflammatory and reparative responses.2, 3 Matrix remodeling takes place at the same time, a process facilitated by matrix metalloproteinases (MMPs) and tissue inhibitors of metalloproteinases (TIMPs) released by HSCs, inflammatory cells, and KCs.4 KC-derived cytokines and unknown mediators may stimulate the expression of MMPs which can degrade the perisinusoidal matrix, allowing migration and proliferation of HSCs to set the stage for scarring.3, 4 Thus, the homeostatic responses initiated by KC-derived mediators underlie the liver's defense and reparative mechanisms against injury.5
Influx of KCs coincides with the appearance of HSC activation markers.6 KCs stimulate matrix synthesis, cell proliferation, and release of retinoids by HSCs, but the mechanisms are still not clearly dissected and likely to be complex as shown by others.3, 7–11 Studies by Matsuoka et al.3 implicated KC-derived TGFβ as responsible for paracrine fibrogenesis. KCs generate ROS, which in turn may enhance HSC activation and collagen I synthesis12 either by their direct actions on the COL1A1 and COL1A2 promoters or by inducing profibrogenic cytokines. KCs also produce NO, which can counterbalance the effects of ROS by reducing HSC proliferation, contractility, and collagen I production; however, NO when reacting with O, generates peroxynitrite (ONOO−), whose potential effects on HSC collagen I production are yet unknown.
Several studies have evaluated the role of conditioned medium from KCs in stimulating HSCs.7, 13 However, to date there is limited information on the intercellular communication between KCs and HSCs, as well as the molecular mechanisms by which KC-derived mediators modulate the fibrogenic response in HSCs. A coculture model was developed to study the role of KCs on HSC activation and collagen I production. This model resembles aspects of the interplay of KCs and HSCs in vivo and has been previously used to gain mechanistic insight on the interaction between hepatocytes and HSCs.12, 14, 15 A potential novel dual mechanism responsible for the increase in collagen I protein in HSCs cocultured with KCs is shown. These effects are mediated in part by transcriptional activation of the COL1A1 and COL1A2 genes. In addition there is decreased turnover of secreted collagen I protein due to low MMP13 activity, the protease that specifically degrades collagen I, coupled to increased expression of TIMP1, an inhibitor of MMP13. While a direct role for oxidative stress may mediate the transcriptional effects; a signaling cascade of ROS (i.e., H2O2), IL-6, and phosphorylated p38 is proposed for regulating effects on matrix degradation.
COL1A1, collagen α1(I) promoter; COL1A2, collagen α2(I) promoter; HSC, hepatic stellate cell; bp, base pair; IL-6, interleukin-6; KC, Kupffer cell; MMP, matrix metalloproteinase; TIMP, tissue inhibitor of metalloproteinase.
Materials and Methods
The coculture model shown in Supplementary Fig. 1 includes a Boyden chamber and a cell culture insert with a polyethylene terephthalate membrane of 8 μm of pore size that allows for diffusion of soluble mediators from the upper to the lower compartment.14 Primary cells are isolated from male Sprague-Dawley retired breeder rats (∼600 ± 25 g) (Charles River Laboratories Inc., Wilmington, MA) by in situ liver perfusion with bacterial pronase and liberase (Roche Diagnostics, Branchburg, NJ), followed by density gradient centrifugation with Histodenz (11% over 17.5%) and elutriation to separate KCs from endothelial cells.14, 16 KCs are collected at a flow rate of 36 mL/minute at 2500 rpm (J2-MC Beckman centrifuge, JE-6B elutriator rotor). Cell viability is determined by trypan blue exclusion. Purity of the HSC fraction (∼95%) is assessed by autofluorescence, and that of the KC fraction (∼96%) by phagocytosis of fluorescein-labeled Staphylococcus aureus and by ED2-positive staining, a specific marker which binds to a membrane antigen found in tissue macrophages.17 KCs can be maintained up to 14 days in culture preserving their ultrastructural and cytochemical characteristics while becoming activated with the time in culture. A range of KC:HSC ratios of 1:1, 2:1, and 3:1 was tested (not shown) and maintained thereafter at ∼3:1, as it caused the most dramatic effects on HSCs and is similar to that found in the liver. To facilitate medium circulation the Boyden chambers were rotated sporadically. In some experiments (Fig. 1; Supplementary Figs. 2 and 3) HSCs were plated on Matrigel (BD Biosciences, San Jose, CA), a basement membrane matrix extracted from the Engelbreth-Holm-Swarm mouse sarcoma. Matrigel is thawed out at 4°C overnight and it is used following the thin gel method (0.5 mm).11, 14 After 1 day of incubation, both the KCs and the HSCs are washed 3 times in serum-free DMEM/F12, and the cell-culture inserts containing the KCs are transferred onto the HSCs. New medium (4 mL without serum) is added to HSCs cultured with KCs or with empty inserts, the latter considered the coculture controls.
Serum contains growth factors and cytokines which may mask potential effects produced by KC-derived mediators on HSCs, or which may obscure the identification of factors released by KCs that affect HSCs. Similarly, serum also contains proteases which eventually degrade secreted procollagen I and collagen I fibers preventing their processing, assembly, and accumulation in the extracellular space. In view of this, all experiments were carried out in the absence of serum, as previously performed by others,18–20 in an effort to specifically target candidate mediators and to more carefully analyze the degradation of secreted collagen I.
was determined by the rate of incorporation of [methyl-3 H]thymidine into the DNA of HSCs as described.14
Immunoblotting and Western Blot Analysis
were performed as previously specified.12, 14, 21 Anticollagen type I Ab (1/5,000) was kindly provided by Dr. Detlef Schuppan (Harvard Medical School, Boston, MA).22, 23 Procollagen type I was detected as several bands, i.e., high MW chains of procollagen α1(I) and α2(I) and the N-terminally processed pCα1(I) and pCα2(I). In blots where the intracellular collagen I expression was analyzed, the band labeled as collagen I corresponds to the pCα1(I) and pCα2(I) which overlap below 200 kDa—the α1 chain is 138.9 kDa and the α2 chain is 129.5 kDa. In the culture media, intact procollagen I predominated along with fully processed collagen I.12, 14, 22, 23 Anti-α-Sma and anti-β-tubulin were obtained from Sigma (St. Louis, MO). Antibodies for IL-6, p38, and phosphorylated p38 were from Santa Cruz Biotechnologies (Santa Cruz, CA). Monoclonal anti-rat MMP13 and anti-rat TIMP1 were from Oncogene (San Diego, CA). The MMP13 antibody recognizes all intermediate-MMP13 (68 kDa), pro-MMP13 (60 kDa), and active MMP13 (48 kDa). Goat anti rabbit IgG and goat anti-mouse IgG (both at 1/5,000) were used as secondary antibodies (Chemicon International, Temecula, CA) and the signal was detected using the ECL system (GE Healthcare-Amersham Biosciences, Piscataway, NJ), and quantified by densitometry.
Northern Blot Analysis and Quantitative Real-Time PCR Analysis.
HSCs were cocultured alone or with KCs for 5 days after which total RNA was isolated using the TRIzol reagent (Life Technologies, Grand Island, NY). In the experiments where mRNA stability was evaluated, 10 ng/mL of actinomycin D were added at 5 days of coculture and samples of RNA were collected at 0, 3, 6, and 12 hours. Northern blot was carried out as described previously.21, 24 Liver RNA was extracted using the RNeasy Mini kit (Qiagen, Chatsworth, CA). All RNA was treated with DNase (Qiagen). RNA (1 μg) was reverse-transcribed using first-strand cDNA synthesis with random primers (Promega, Madison, WI). Quantitative real-time PCR was done with the following PCR primers on an ABI Prism 7900HT sequence detection system (Applied Biosystems, Foster City, CA): COL1A1 forward, 5′-CCT CAA GGT TTC CAA GGA CC-3′ and COL1A1 reverse, 5′-CAA TCC ATC CAG ACC GTT GTG-3′; COL1A2 forward 5′-CCT CAA GGT TTC CAA GGA CC-3′ and COL1A2 reverse 5′-CAA TCC ATC CAG ACC GTT GTG-3′. All values were normalized to GAPDH.
Nuclear In Vitro Run-on Transcription Assays
were carried out using nuclei from HSCs cultured alone or with KCs for 5 days as described.21, 24
Primary HSCs and KCs were isolated from transgenic mice harboring the −17 bp to +54 bp of the promoter of the mouse COL1A2 gene cloned upstream of the E. coli β-Gal reporter gene (LacZ). These transgenic mice were obtained from Dr. Benoit de Crombrugghe (M.D. Anderson Cancer Center, Houston, TX).25–29 HSCs were cultured alone or with KCs for 5 days and β-gal activity was measured in HSCs using a chemiluminiscent reporter assay (Galactolight Plus, Tropix, Bedford, MA). Results are expressed as U/mg of protein.
Transfection Experiments and Reporter Assays.
Reporter DNA constructs containing upstream sequences of the human COL1A1 and COL1A2 promoter linked to the Luc gene were provided by Dr. Vincent Falanga (Boston University School of Medicine, Boston, MA)30 and Dr. Francesco Ramírez (Hospital for Special Surgery, New York, NY),31 respectively. In these constructs, the human COL1A1 promoter sequences span from −2500 bp to +113 bp with the following 5′ endpoints: −2500, −804, −515, −265, −232, −209, −190, −174, and −72 bp.30 The human COL1A2 promoter sequences extend from −3500 bp to +58 bp with 5′ endpoints of −3500, −772, −378, −183, and −108 bp.31 The −1168 bp human IL-6 reporter vector was a generous gift from Drs. Haegeman and Vanden Berghe (University of Gent-VIB, Belgium).32 Parallel transfection of the corresponding empty vectors pGL3-Luc, PXP1-Luc, and P-Luc at equivalent concentrations was performed. Total amount of plasmid DNA was equalized using pBluescript SK- (Stratagene, La Jolla, CA). Primary HSCs (after 3 to 4 passages to increase transfection efficiency) were plated at a density of 150,000 cells/well in 6-well plates and recently isolated KCs were plated at a density of 450,000 cells/insert. Complexes containing Fugene 6 (Roche Molecular Biochemicals, Indianapolis, IN) plus plasmid DNA were prepared with a final concentration of plasmid DNA for each of the chimeric constructs of 1 μg/mL. Parallel cotransfection with 25 ng/mL of the control pRL-null (Promega, Madison, WI) containing the cDNA encoding for Renilla Luc was performed to normalize for transfection efficiency. HSCs were incubated in the presence of the transfection mix for 48 hours, after which the medium was replaced and the inserts containing the KCs were transferred onto the HSC plates. In some experiments cells were treated with 10 μmol/L H2O2 in the presence of 100 U/mL of catalase or of boiled catalase (Fig. 6C,D and Fig. 7C,D). Samples for the Luc activity were collected and the reaction was run using a kit from Promega as described.33
Pulse and Pulse-Chase Analysis
were carried out as in earlier publications12, 15 using either 50 μg of HSC lysate protein or 100 μg of protein from the culture medium. Samples were pulsed at 5 days and intracellular collagen I synthesis was followed up to 12 hours. For the pulse-chase experiment, samples were pulsed for 12 hours and then chased with unlabeled methionine for the following 12 hours. Samples were immunoprecipitated with anticollagen I IgG, electrophoresed in a 5% SDS-PAGE, dried, and autoradiographed. To evaluate for the rate of total protein synthesis or total protein degradation, proteins from HSC lysates or culture medium were TCA-precipitated, the supernatant removed, the precipitate resuspended, and the TCA-precipitated counts quantified. The t1/2 of collagen I protein was estimated from the semilogarithmic plot.
was done to detect latent, intermediate, and active forms of MMP2 and MMP9. The medium was mixed with Laemmli's loading buffer in the absence of reducing agents and resolved in a 7.5% PAGE with 0.1% (wt/vol) gelatin, a substrate that can be cleaved by MMP2 and MMP9. SDS in the gel was removed by incubation with 2.5% Triton X-100 for 1 hour. Proteolytic activities were developed in 5 mmol/L CaCl2, 150 mmol/L NaCl, and 50 mmol/L Tris (pH 7.5) overnight at 37°C, stained with Coomassie blue R-250 in 10% acetic acid/40% methanol, and distained in 40% methanol/10% acetic acid/3% glycerol. MMPs activity was visualized as bands of lysis (pro-MMP9 is ∼92 kDa, active MMP9 is ∼82 kDa, intermediate MMP2 is ∼72 kDa, pro-MMP2 is ∼68 kDa, and active MMP2 is ∼62 kDa), which appear clear in a dark background.34
Intracellular and extracellular H2O2 were assessed using the fluorescent probe DCF-DA (Molecular Probes, Eugene, OR)21 and the FOX method,35 respectively. IL-6 concentration in the culture medium was determined using an ELISA kit from Alpco Diagnostics (Windham, NH). Viability was assessed by the MTT assay. Recombinant rat IL-6 was purchased from Sigma. siRNA-IL-6 and siRNA-control (a nontargeting siRNA) were transfected onto KCs only and prior to setting up the coculture by using the siRNA transfection reagent according to the manufacturer's protocol (Ambion, Austin, TX). For the catalase protection studies, cells were infected with an adenovirus containing the cDNA encoding for catalase or a Null adenovirus at m.o.i. = 100. The adenovirus were provided by Dr. Arthur I. Cederbaum (Mount Sinai School of Medicine, New York, NY).36 Catalase activity was assayed 48 hours after infection according to the method of Claiborne.37
Experiments were analyzed by Student t test. In all the experiments there were at least n = 3. The statistical significance can be found in the legend to each figure.
KCs Promote Phenotypic Changes in HSC.
Preliminary experiments were performed using direct coculture of KCs and HSCs to determine if cell-cell contact was necessary to activate HSCs and stimulate collagen I production; however, effects were identical (not shown) to those described below using Boyden chambers. Experiments were then carried out to evaluate the feasibility of our model, studying the effect of time of coculture with KCs on HSC morphology, phenotype, and induction of parameters of HSC activation.
The phenotypic changes in HSCs cocultured with KCs were studied from 1 to 6 days when HSCs were seeded on polystyrene plates, and from 1 to 12 days when HSCs were plated on Matrigel. Stretching, nuclear and cellular enlargement, cytoplasm spreading, elongation of processes establishing contacts among cells, loss of lipid droplets and vitamin A were more apparent in HSCs cocultured with KCs (Supplementary Fig. 2, panels labeled as “coculture”). Cocultured HSCs plated on Matrigel extended cellular processes or filopodia which adhered to and extended along matrix fibers14 (Supplementary Fig. 2, right panels).
KCs Induce HSC Proliferation.
HSC proliferation was assessed by the rate of incorporation of [methyl-3 H]thymidine into the DNA. A time-course experiment was carried out with HSCs cultured alone or with KCs grown on polystyrene plates (up to 7 days), or on Matrigel-coated plates (up to 12 days). Consistent with the light micrographs (Supplementary Fig. 2, a 4-fold increase in the proliferation of HSCs cultured with KCs compared to HSCs cocultured with empty inserts was found under both culture conditions (Fig. 1).
HSC Activation is Modulated by KCs.
Under the same experimental conditions of Fig. 1, a Western blot analysis for α-Sma was carried out. There was a greater time-dependent increase in α-Sma expression in HSCs cocultured with KCs when compared to HSCs cultured alone (Supplementary Fig. 3) which correlated the phenotypic changes and the proliferation results depicted in Fig. 1 and Supplmentary Fig. 2. Analogous results were obtained when cells were plated on Matrigel; however, longer coculture times (from 6 to 12 days) were necessary for the changes in α-Sma to occur (Supplementary Fig. 3, right panel). For the purpose of simplicity, all the experiments described below were performed on polystyrene dishes, because Matrigel elicited similar effects on HSCs albeit at later time.
COL1A1 and COL1A2: Targets for KC-Derived Mediators.
KCs increased both COL1A1 and COL1A2 mRNA levels about 3.9-fold and 2.8-fold, respectively in cocultured HSCs when compared to HSCs, while the GAPDH mRNA signal remained unchanged as assessed by Northern blot analysis and real-time PCR (Fig. 2A). Whether COL1A1 and COL1A2 mRNA stability was modified by the coculture was next examined, and found that the t1/2 of both COL1A1 and COL1A2 mRNA was similar for HSC cultured alone or with KC (∼5.5-6 hours) (Fig. 2B). The enhanced COL1A1 and COL1A2 expression induced by KCs occurred most likely through a transcriptional mechanism (Fig. 2C). Newly transcribed COL1A1 and COL1A2 mRNA increased 3.1-fold and 2.4-fold, respectively, in HSCs cocultured with KCs compared to HSCs cultured alone (Fig. 2C).
KC-Derived Mediators Transactivate COL1A1 and COL1A2 Promoters.
To further explore the effects of KCs on collagen I transcription, HSCs from transgenic mice for the COL1A2 promoter linked to the β-gal reporter gene were isolated and cocultured with control KCs for 5 days. β-Gal activity increased ∼4.8-fold in HSCs in coculture with KCs, indicating transactivation of the COL1A2 promoter (Fig. 2D). To identify the minimal region in both promoters that confers basal and induced transcriptional activity in the cocultures, HSCs were transfected with a series of deletion constructs of both promoters before starting the coculture (Fig. 3). The minimal sequences of the COL1A1 and COL1A2 promoters required for increased basal and KC-induced activation were localized to the −515 bp to +113 bp region of the COL1A1 promoter (Fig. 3A) and to the −378 to +58 bp region of the COL1A2 promoter (Fig. 3B).
KCs Decrease the Turnover of Extracellular Collagen I From HSCs.
To evaluate for the rate of collagen I protein synthesis, a pulse analysis was carried out. The results revealed a constant 2.7-fold difference at all time points in collagen I expression in HSCs cocultured with KCs compared to HSCs cultured alone (Fig. 4A,B), whereas total protein synthesis was similar in both culture settings (Fig. 4C). Turnover of secreted collagen I, as studied by pulse-chase analysis, was reduced in HSCs in coculture with KCs when compared to HSCs (Fig. 4D). The half-life of secreted collagen I was estimated to be ∼2.8 hours for HSCs cultured alone and ∼7 hours for HSCs in coculture (Fig. 4F), while total secreted protein degradation was alike for both culture systems (Fig. 4G). The half-life of intracellular collagen I was similar in both culture settings (Fig. 4E) along with similar rate of intracellular protein degradation (not shown).
Intracellular and Extracellular Collagen I and TIMP1 Proteins Are Upregulated in HSCs in Coculture With KCs.
A Western blot analysis was carried out in HSC lysates and culture medium at selected time intervals (Fig. 5A). A more evident time-dependent increase in intracellular and extracellular collagen I was observed in HSCs in coculture with KCs, compared to HSCs cultured alone. Because collagen I appeared to accumulate more in the culture medium of HSCs cocultured with KCs (Fig. 5A, lower blot), the expression of several MMPs and TIMP1, known to participate in ECM remodeling, was analyzed. Samples of medium were evaluated for the gelatinolytic activity of MMP2 and MMP9 (Fig. 5B) and the expression of MMP13 and TIMP1 (Fig. 5C). Gelatin zymography of HSCs cultured alone or with KCs showed hardly any detectable active MMP2 or MMP9 (Fig. 5B); however, in the cocultures both pro-MMPs (which do have gelatinolytic activity but not collagenolytic activity) were highly expressed, possibly due to the presence of a modest amount of TGFβ in the medium (not shown); active MMP2 and MMP9 were barely detected. Western blot analysis illustrated about a 2.2-fold increase in TIMP1 expression in the cocultures compared to HSCs, which may account for an associated downregulation of active MMP13 (Fig. 5C), a metalloproteinase that specifically degrades extracellular collagen I. TIMP2 and MT1-MMP were not detected (not shown).
H2O2, a Candidate Mediator for Collagen I Induction by KCs.
Intracellular ROS, mainly H2O2, were elevated by 50% in HSCs cocultured with KCs compared to HSCs cultured alone (Fig. 6A). There was a time-dependent increase in the amount of ROS, mainly H2O2, in the medium of HSCs in coculture with KCs. This increase was higher than for HSCs cultured alone from 3 days on (Fig. 6B). Similar antioxidant defense was found in HSCs cultured alone or with KCs (not shown). To test whether the increase in ROS could elevate collagen I, experiments were carried out in the presence or absence of 100 U/mL of catalase, which decomposes H2O2, or by infection with either a control adenovirus (Ad-Null) or an adenovirus containing the cDNA encoding for catalase.36 After 5 days of coculture, proteins from HSCs and culture medium were immunoblotted. Catalase treatment and infection with the adenovirus containing the cDNA encoding for catalase resulted in a decrease in intracellular and extracellular collagen I to almost basal levels (Fig. 6C). Catalase activity increased in HSCs infected with the adenovirus from 1.1 ± 0.1 to 12.1 ± 1.2 U/mg of protein and in KCs from 3.4 ± 0.4 to 37.6 ± 2.1 U/mg of protein. Overexpression of catalase decreased intracellular ROS from 81 ± 1.1 to 8 ± 0.3 AU/mg of protein in HSCs and from 127 ± 1.9 to 6 ± 0.2 AU/mg of protein in HSCs in coculture as well as from 187 ± 9.3 to 16 ± 1.1 AU/mg of protein in KCs in coculture, and extracellular ROS decreased from 95 ± 5.7 to 5 ± 0.4 nM in HSC cultured alone and from 175 ± 10.3 to 10 ± 0.8 nM in the coculture. Catalase treatment decreased ROS levels by 90% (not shown). Cell viability under catalase treatment was monitored by the MTT assay and found to be 100%.
To dissect whether the protective role of catalase was mediated by targeting the induction of the COL1A1 and COL1A2 promoters, the activity of both promoters was tested in the presence of added catalase. Catalase prevented both the basal as well as the increased responsiveness of the COL1A1 and COL1A2 promoters by KCs, validating a role for H2O2 (Fig. 6D).
H2O2 Induces IL-6 Production, Which Through a p38 Phosphorylation-Dependent Mechanism, Elevates TIMP1, Lowers MMP13, and Decreases the Turnover of Extracellular Collagen I.
There was a time-dependent upregulation of IL-6 levels in the coculture medium which was greater to that found in the medium from HSCs cultured alone (Fig. 7A). To assess the responsiveness of HSCs to IL-6, HSCs were cultured alone or in the presence of recombinant rat IL-6 and the culture medium was analyzed by Western blot. Treatment with IL-6 elevated extracellular collagen I 4.1-fold and TIMP1 2.9-fold while active MMP13 disappeared (Fig. 7B). HSC transfected with the COL1A1 or the COL1A2 promoter constructs and treated with IL-6 for 6 and 24 hours showed no significant increase in transactivation of either promoter (Fig. 7C). To determine whether the upregulation of IL-6 shown in Fig. 7A could be mediated, at least in part, by H2O2, and to identify the potential source of IL-6, either HSCs or KCs were incubated with 25 μmol/L H2O2, with or without catalase, and the expression of IL-6 was analyzed by Western blot (Fig. 7D). IL-6 was increased in both KCs and HSCs in the presence of H2O2, and catalase prevented this induction, establishing a potential link between ROS (e.g., H2O2) and IL-6 (Fig. 7D). This relationship was further tested by IL-6 promoter studies (Fig. 7E), which indicated IL-6 promoter-induction by H2O2 mainly in KCs. IL-6 levels were considerably high in the culture medium of KCs and in the cocultures, while they were modest in HSCs cultured alone (Fig. 8A). Active MMP13 was present only in HSCs cultured alone and TIMP1 was elevated in HSCs cultured alone or with KCs; however, KCs expressed low TIMP1 (Fig. 8A).
To investigate potential signals between the H2O2-mediated IL-6 induction and TIMP1 upregulation, the phosphorylation status of several stress activated kinases was studied. HSCs in coculture demonstrated about a 2.7-fold increase in the phosphorylation of p38 when compared to HSCs cultured alone (Fig. 8B) but not of ERK or activation of PI3K (not shown). Incubation with SB203580, an inhibitor of p38 phosphorylation, but not with PD98059, an inhibitor of ERK phosphorylation, prevented the increase in TIMP1 expression by the cocultures, without affecting IL-6 levels (Fig. 8C), suggesting a link between IL-6, phosphorylated p38, and TIMP1. To establish a connection between the levels of H2O2, IL-6, phosphorylated p38, TIMP1, active MMP13, and extracellular collagen I, cells were incubated with catalase, a neutralizing antibody to IL-6, or KCs were transfected with either control-siRNA or with IL-6-siRNA before transferring the inserts onto HSCs. Extracellular collagen I decreased under both treatments (catalase and anti-IL-6), while active MMP13 increased, and both TIMP1 and phosphorylated p38 decreased in the culture medium of HSCs cocultured with KCs (Fig. 8D). Similarly, transfection of KC with siRNA-IL-6 prevented the increase in collagen by the cocultures pointing also at KC-derived IL-6 as a likely candidate mediator. These results suggest a direct link between H2O2, IL-6 (mostly coming from KCs), phosphorylated p38, TIMP1, MMP13, and extracellular collagen I which may account, at least in part, for the KC-mediated increase in collagen I accumulation in the culture medium of HSCs.
ROS, cytokines, and growth factors are increasingly considered potential mediators of HSC activation; however, much of the evidence is indirect and the current in vitro approaches do not allow for sustained generation of ROS intracellularly in order to identify regulatory pathways or intracellular mediators. To overcome these obstacles, we developed a coculture system to characterize the intercellular communication between KCs and HSCs, and to better understand the mechanisms by which KC-derived mediators induce a fibrogenic response in HSCs. This system with constant generation of the potential mediators may be more reflective of the physiological conditions in the liver, and allows (1) identification of diffusible mediators responsible for modulating collagen I production; (2) permits evaluation of the role of KC-derived mediators on extracellular matrix production; and (3) may help to explore molecular mechanisms underlying the increase of collagen type I protein.
Because HSC activation is characterized by alterations in cellular morphology,1 the potential phenotypic changes mediated by KCs were first analyzed. A more activated phenotype was found in HSCs cocultured with KCs compared to HSCs cultured alone. The features observed in the presence of KCs, e.g., spreading and elongation, generation of cytoplasmic processes, loss of vitamin A content and of lipid droplets, may be quite relevant to HSCs biology as they simulate in vivo events in the liver (e.g., HSCs migration and bridging fibrosis). Indeed, Matsuoka et al. have shown that KCs modulate HSCs phenotype and behavior by TGFβ3 and that a high-fat diet sensitizes HSCs to the stimulatory effects of KC-derived factors.10 This finding, along with previous work by others,3, 7–10 reinforced the working hypothesis that KC-derived mediators may regulate HSCs biology.
Given that features of perpetuation of HSCs activation include increased cell accumulation from proliferation and directed migration, cell growth was analyzed, and observed greater proliferation for HSCs in coculture with KCs. Similar results were obtained for cells grown on polystyrene plates and on Matrigel-coated plates, which maintains the cells in a more quiescent state,38 indicating that even when HSCs are more quiescent, KCs release mediators in quantities large enough to stimulate HSCs proliferation.14
Friedman and Arthur demonstrated that activation of cultured rat HSCs by KCs conditioned medium enhanced matrix synthesis and stimulation of cell proliferation via induction of PDGF receptors.7 Gressner et al. described that KCs induce the synthesis and secretion of proteoglycans and hyaluronate by HSCs in culture.8, 9 A major goal of this study was to assess whether KC-derived mediators could lead to activation of HSCs and matrix deposition and to dissect other potential molecular mechanisms involved in connection with ROS and cytokines. The cytoskeleton protein α-Sma, typically considered a marker for HSCs activation, was elevated in HSC in coculture either on polystyrene plates or on Matrigel-coated plates when compared to HSCs cultured alone. Together, these data revealed a clear effect by KCs on HSCs which led to studies examining collagen I, a major component of pathological ECM.
To address this issue, COL1A1 and COL1A2 mRNA levels were studied and found to be elevated in HSCs in coculture compared to HSCs cultured alone. Increased mRNA levels could be a reflection of elevated mRNA stability. Indeed, posttranscriptional regulation of COL1A1 mRNA has been shown to be regulated in HSCs cultured alone by the conserved stem-loop structure at the 5′-end and by the interaction between αCP(2) and the 3′-untranslated region.39–42 The half-life of both COL1A1 and COL1A2 mRNA, however, remained similar under both culture conditions. These results suggested a first KC-mediated profibrogenic mechanism in HSCs by increasing COL1A1 and COL1A2 mRNA without altering mRNA stability.
Nuclear in vitro transcription run-on assays further confirmed that the KC-mediated effect on COL1A1 and COL1A2 mRNA involved transcriptional control of both genes. In addition, primary HSCs from transgenic mice harboring the COL1A2 promoter linked to the β-gal reporter gene and cocultured with control KCs showed elevated β-gal activity compared to HSCs cultured alone, indicating transactivation of the COL1A2 promoter. Transfection experiments narrowed down the minimal region of the COL1A1 and COL1A2 promoters conferring basal and KC-induced responsiveness located below −515 bp for the COL1A1 promoter and below −378 bp for the COL1A2 promoter, respectively. These 2 regions have been previously described by others to be sensitive to acetaldehyde, ROS, TGFβ, and TNFα (in the COL1A1 promoter)18, 43, 44 and to acetaldehyde and ROS (in the COL1A2 promoter).24, 45 All these mediators play a pivotal role in the development of liver disease and a potential synergistic loop has even been proposed among some of them.18, 43, 46 The activity of the −772COL1A2 construct was significantly lower than that of the −3500COL1A2 and −378COL1A2 constructs given the presence of a silencer element between −772 and −378 bp which has been described by others.47
Based on the above, a fibrogenic response (i.e., collagen I protein deposition) appeared likely. There was a time-dependent increase in intracellular and extracellular collagen I expression which was more apparent for HSCs cocultured with KCs. Gel zymography of HSCs cocultured with KCs indicated no major change in the activity of MMP2, with potential collagenolytic activity, or in MMP9, which can cleave collagen I but can also activate latent TGFβ, thereby inducing collagen I. Failure to degrade the accumulated scar matrix is a likely reason why fibrosis evolves. Progressive fibrosis may be associated with marked increases in TIMP1 and TIMP2, leading to a net decrease in protease activity and therefore greater matrix accumulation. HSCs are the major source of these inhibitors.48 MMP13, the main protease that specifically cleaves collagen I, decreased in the cocultures while TIMP1, the inhibitor of MMP13, increased, suggesting a second KC-mediated profibrogenic mechanism, based on collagen I accumulation in the extracellular space.
The turnover of secreted collagen I was assessed to determine if increased levels of collagen I in the culture medium may also reflect increased accumulation of the secreted protein. There was a constant time-dependent increase in intracellular collagen I levels under both culture conditions while the half-life of secreted collagen I was higher for HSCs cocultured with KCs compared to HSCs cultured alone. Intracellular degradation of procollagen I occurs mostly via the ubiquitin-proteasome pathway49 while breakdown of extracellular procollagen I and collagen I usually occurs via a MMP-mediated mechanism.50 The above data indicated that a potential mechanism to account for the differences in collagen I expression may be increased accumulation of pro-collagen I and collagen I proteins in the culture medium of HSCs coincubated with KCs. The possibility of decreased degradation of collagen I in HSCs by KC-derived mediators was then considered.
Fibrillar collagen type I provides the liver's tensile strength, and only MMPs can degrade it in the extracellular space. Alterations in the balance between matrix deposition and degradation brought about by changes in the activities of MMPs and TIMPs may contribute significantly to liver disease and fibrosis.48 Previous work by Han et al. has described an essential role for MMPs in IL-1-induced myofibroblastic activation of HSCs on collagen substratum.11 It is likely that matrix remodeling may be affected by cytokines many of which remain sequestered within matrix fibers where they can be released by HSCs, KCs, or both.51 Because IL-6 levels increased in a time-dependent fashion in the coculture medium, the question of whether IL-6, by inducing TIMP1 expression, and acting through an additional target, i.e., MMP13, could counteract collagen I proteolysis was raised. HSCs cultured alone appeared responsive to added IL-6, increasing extracellular collagen I and TIMP1, and decreasing MMP13.
Greenwel et al.43 have shown a direct link between COL1A1 induction and H2O2. Both intracellular and extracellular H2O2 were elevated in HSCs cocultured with KCs compared to HSCs cultured alone. To determine if H2O2 was a candidate mediator for the effects of KCs on HSCs, the possibility that catalase could prevent the increase in collagen I was then evaluated. Catalase both prevented the accumulation of collagen I in the culture medium of HSCs coincubated with KCs, and blocked the basal and KC-induced transactivation of the COL1A1 and COL1A2 promoters, validating a role for H2O2 in mediating the effects of KCs on HSCs collagen I production. Because H2O2 levels were also elevated in the cocultures, the possibility that the increase in IL-6 could be mediated by an H2O2-dependent mechanism was then investigated. Both KCs and HSCs cultured alone were responsive to H2O2, upregulating IL-6; however, the response appeared to be greater in KCs than in HSCs, and it was blocked by catalase. Both the culture medium from KCs cultured alone or with HSCs expressed equal amounts of IL-6, pointing to KCs as the main source of this cytokine, which was latter validated using siRNA-IL-6, although a role for HSC-produced IL-6 cannot be ruled out.
A possible upstream signal, i.e., a stress-sensitive kinase, as a link between IL-6 and TIMP1 was considered. Of the kinases analyzed, only phosphorylated p38 was up-regulated in HSCs in coculture with KCs. Addition of SB203580 but not of PD98059, blocked the TIMP1 increase, suggesting that phosphorylation of p38 was the switch upregulating TIMP1 expression. To establish a connection between H2O2, IL-6, phosphorylation of p38, TIMP1, active MMP13, and extracellular collagen I deposition, HSC alone or in coculture were treated with catalase or a neutralizing antibody to IL-6. Catalase and the neutralizing antibody to IL-6 reduced extracellular collagen I possibly via reprogramming of TIMP1 and MMP13 activity in the coculture setting. Indeed, phosphorylation of p38 and TIMP1 expression were down-regulated by both treatments, while active MMP13 was up-regulated, and consequently extracellular collagen I was reduced. Moreover, transfection of an siRNA-IL-6 onto KCs before the coculture setting caused similar effects suggesting a major role for KC-derived IL-6 in the fibrogenic response of HSCs to KCs.
In summary, at least two novel regulatory mechanisms may be involved in the fibrogenic response of HSCs to KCs. First, H2O2 may participate in modulating the transactivation of both COL1A1 and COL1A2 promoters; and second, an intracellular plus an extracellular network whereby H2O2-induction (in KCs and/or HSCs) of IL-6 (mostly in KCs) could drive extracellular collagen I degradation may coexist. The data support that p38 phosphorylation is a key regulatory point in avoiding the generation of a matrix degrading phenotype by controlling TIMP1 and MMP13 activity.
I am very grateful to Dr. Arthur I. Cederbaum for his generous help in analyzing and discussing the data, and to Dr. Scott L. Friedman for his careful review of the manuscript.