Essential Role of Matrix Metalloproteinases in Interleukin-1-induced Myofibroblastic Activation of Hepatic Stellate Cell in Collagen^*

Isolation Rat HSC

HSC were isolated from male Wistar rats (Charles River) by the Non-Parenchymal Liver Cell Core of the Research Center for Alcoholic Liver and Pancreatic Diseases. The normal Wistar rats were treated by in situ sequential digestion with Pronase and collagenase followed by arabinogalactan gradient ultracentrifugation as described previously (10, 44). Briefly, the liver was sequentially digested with Pronase (Roche Applied Science) and type IV collagenase (Sigma) by in situ perfusion. Parenchymal cells were removed by centrifugation of the digest at 50 × g for 2 min, and nonparenchymal cells were laid on top of the four density gradients of arabinogalactan (Sigma). The gradients were centrifuged at 21,400 rpm for 40 min at 25 °C using a Beckman SW41 Ti rotor (Beckman Instruments). A pure fraction of HSC was recovered from the interface between the medium and the density of 1.035. The purity of HSC was assessed by phase-contrast microscopy and ultraviolet excited fluorescence microscope, and the viability was examined by the trypan blue exclusion test. Both the purity and viability of the cells always exceeded 95%.

Cell Culture and Embedding in Three-dimensional Matrix

The model used in the most of the experiments described below is based on a comparison of morphologic and biochemical parameters of HSC cultured in three-dimensional matrices (type I collagen gel, Matrigel, fibrin) and those on plastic. Freshly isolated HSC were plated on 10-cm culture dish with 10% FBS in DMEM for 2 days. During this period, HSC retained the quiescent state as monitored by star-shaped morphology with long, prominent dendritic processes and autofluorescence of vitamin A-containing droplets. Then cells were detached by a trypsin/EDTA treatment followed by neutralization by 10% FBS in DMEM. After washing with serum-free DMEM the cells were suspended in 4 ml of DMEM. The cell density was adjusted to 2 × 10⁶/ml. To plate the cells as a monolayer on plastic, 2 ml of the cell suspension was mixed with 18 ml of DMEM containing 10% FBS and seeded of 1 ml in each well of a 24-well plate. Another 2-ml cell suspension was mixed with type I collagen gel mixture as described previously (45). Briefly, this mixture was prepared by mixing 0.6 ml of 10× minimum essential medium, 0.6 ml of 0.1 N NaOH, 1.4 ml of DMEM, and 4.8 ml of Vitrogen (Cohesion Technologies). An aliquot of 0.4 ml was loaded in the wells of a 24-well plate and incubated at 37 °C for 2 h. After forming of lattices, 1 ml of DMEM and 10% FBS were added on top and cultured in 5% CO₂ at 37 °C for 6 h. Then the monolayers or lattices were washed gently by DMEM followed by culturing in 1 ml of DMEM (0.5% FBS) with cytokines (R&D Systems) as mentioned in the text. To make the fibrin clot, a stock was prepared by mixing 3 ml of 10 mg/ml fibrinogen in DMEM and 1 ml of cell suspension. Aliquots of 5 μl of thrombin (250 units/ml) were preloaded in the wells followed by adding 0.4 ml of the fibrinogen cell mixture. To make Matrigel gel lattices, the stock (BD Biosciences) was diluted twice with cold DMEM followed by mixing with the cell suspension. For some experiments, a sandwich gel was made by placing an additional 0.2-ml matrix layer before loading the collagen lattices on the top. For inhibition experiments, purified recombinant TIMP-1 (CC3328, Chemicon) was applied to the culture.

Deprivation of MMP-9 during IL-1α-induced Activation of HSC

Isolated rat HSC were embedded into three-dimensional type I collagen in the bottom chamber of a Transwell (BD Biosciences). 400 μl of gelatin-conjugated Sepharose 4B (Amersham Biosciences) or the control beads were loaded into the insert chambers. After culture with 20 ng/ml IL-1α for 6 days the cells were fixed for immunostaining of α-SMA. The conditioned medium was examined for gelatinase activities.

Zymography

Prior to gel electrophoresis the conditioned medium was briefly centrifuged and mixed with a SDS-PAGE sample buffer in the absence of reducing agents. The conditioned medium was resolved by 10% polyacrylamide separation gel with 0.1% (w/v) gelatin (Sigma). To prepare the casein zymography, the protein was dissolved in 0.01 N NaOH before casting into the gel. Electrophoresis was carried out at 4 °C with 120 volts for 16 h. After electrophoresis, SDS in the gel was deprived by incubation with 2.5% Triton X-100. Proteolytic activities were developed in a buffer containing 5 mM CaCl₂, 150 mM NaCl, and 50 mM Tris, pH 7.5, for 16 h at 37 °C and visualized by staining with Coomassie Blue R-250. For most of the experiments, 35 μl of the original conditioned medium was resolved by zymogram. To examine the weak activity of MMP-9 from the fully activated HSC (myofibroblast-like HSC), gelatinases from 300 μl of conditioned medium were enriched by pull-down with gelatin-conjugated Sepharose 4B prior to zymography.

Western Blot Analysis

Because some components in fetal bovine serum were found to be tightly associated with the collagen matrix and recognized by the secondary antibody in Western blot analysis, quiescent HSC was cultured in serum-free DMEM, and a minimal concentration of PDGF at 0.2 ng/ml was added in the basal medium for culturing fully activated HSC to prevent apoptosis. The conditioned medium was resolved by 12.5% SDS-PAGE, and the protein was transferred to Immobilon-P (Millipore). After blotting with 10% nonfat milk, the individual membranes were probed with either anti-rat MMP-13 (mAB13426, Chemicon), MMP-3 (AB19150, Chemicon), TIMP-1 (MS-570, NeoMarkers), or MMP-9 (AB805, Chemicon) overnight. After incubation with horseradish peroxidase-conjugated secondary antibodies, the membranes were developed by Super Signal West Pico Chemiluminescent substrate (Pierce).

MMP-9 Activity Assay

The fluorogenic peptide substrate, DNP-Pro-Leu-Gly-Met-Trp-Ser-Arg (Calbiochem), was dissolved in 15% dimethyl sulfoxide. Assays were carried out by incubation of 5 μl of conditioned medium, 45 μl of H₂O, 50 μl of 36 μM peptide dilution, and 50 μl of buffer containing 100 mM NaCl, 30 mM CaCl₂ in 50 mM Tris, pH 7.4, at 37 °C for 4 h. The hydrolysis of the peptide was measured by the fluorescence at λ_ex = 280 nm and λ_em = 360 nm in a Hitachi F-2000 spectrofluorometer.

Immunohistological Staining

After cultivation the HSC in matrix lattices and on plastic were washed by phosphate-buffered saline and fixed in cold methanol for 20 min for three-dimensional gel and 10 min for monolayers. After three washes the cells were permeabilized by incubation with cold 0.15% Triton X-100 with 0.1% bovine serum albumin (fraction V from Sigma) in phosphate-buffered saline for 15 min. After treatment with blocking solution containing 5% bovine serum albumin for 60 min, the staining was performed by incubation of 1:200 dilution of fluorescein isothiocyanate-conjugated monoclonal antibody against α-smooth muscle actin (α-SMA, Sigma) in phosphate-buffered saline with 5% bovine serum albumin fro 16 h at 4 °C. After three washes, the nuclei were stained by 0.1 μg/ml DAPI for 10 min. For some experiments the colorimetric staining with horseradish peroxidase-conjugated secondary antibodies and AEC substrates (Labvision) were used. The cells were viewed using a Nikon microscope equipped with digital camera available at the Imaging Core of the USC Research Center for Liver Diseases. To examine the colocalization of MMP-9 and α-SMA, a dual immunofluorescence staining was performed onto human fibrotic liver biopsies by the Morphology Core of the Research Center for Alcoholic Liver and Pancreatic Diseases with approved IRB protocol. Briefly, a paraffin block was sectioned at a thickness of 5 μm followed by a dewaxing process. After hydration, the slides were treated with 0.1% trypsin at 37 °C for 30 min and then blocked by a solution containing 5% skim milk (BD Biosciences), 0.25% serum, and 1% bovine serum albumin (EM Science). The slides were incubated with rabbit antibodies for human MMP-9 (AB805, Chemicon) (1:300) and mouse monoclonal antibody for α-SMA (1:300) in the blocking solution for 16 h at 4 °C. After washing of the slides the fluorescein isothiocyanate-conjugated anti-mouse IgG and Cy3-conjugated anti-rabbit IgG were applied at a dilution of 1:400. Before mounting, the slides were stained with 1 μg/ml DAPI.

Induction and Activation of MMP-9 during IL-1α-induced Activation of HSC in ECMs

A general design of this study was to establish a three-dimensional culture of HSC by embedding isolated quiescent HSC in ECM matrix and to examine cytokine regulation of MMPs and HSC activation. We first tested type I collagen gel because this model was shown to recapitulate best the quiescent morphology of HSC in vivo (31). HSC were isolated from rats and plated on a plastic dish with medium containing 10% FBS for 2 days. More than 95% of the cells retained the quiescent phenotype of HSC as judged by perinuclear retinoid droplets and star-shaped morphology with prominent dendritic protrusion. Then the cells were either reseeded on plastic as a monolayer or embedded in the reconstituted fibrillar type I collagen. The cells were cultured in basal medium with an increasing concentration of IL-1α. After an additional 6 days of culture in the absence of IL-1α, the cells in three-dimensional collagen retained a quiescent state with intracellular vitamin A droplets and the absence of α-SMA immunoreactivity (Fig. 1A). Treatment of HSC with IL-1α induced apparent trans-differentiation of the quiescent HSC into myofibroblast-like cells with the formation of α-SMA-positive fibers. The initial induction of α-SMA for monolayer and three-dimensional-collagen was evident at 0.2 and 2 ng/ml, respectively. At the concentration of 20 ng/ml, IL-1α brought formation of heavy bundles of stress fiber in the cells.

Gelatinolytic activities in the conditioned medium were resolved and are shown in Fig. 1B. To our surprise, robust gelatinase activities with a molecular mass of 105–95-kDa were produced by IL-1α-treated HSC in three-dimensional collagen. In contrast, IL-1α or three-dimensional collagen matrix alone was without effect. The 105-kDa gelatinase represents the rat pro-MMP-9 with a prodomain of ~100 amino acid residues. The 95-kDa gelatinase is regarded as the active form of rat MMP-9 generated by the cleavage of the prodomain. In fact, only pro-MMP-9 was generated when the cells were treated with a low concentration of IL-1α (0.2 ng/ml). Further, after a continued culture for 6 days with IL-1α, the collagen gel was completely degraded, suggesting that fibrillar collagenase activities were induced by IL-1α (Fig. 1C). We also assessed serine proteinase activities by casein zymography. As shown in Fig. 1D, a major serine proteinase with an approximate molecular size of 75-kDa was expressed at the basal state in the collagen lattice, and the treatment with IL-1α induced the generation of a lower molecular mass serine proteinase, presumably through an activation process.

To address the concern that cellular contaminants such as Kupffer cells may provide the source of MMP-9, we performed detailed histological analysis of the three-dimensional cultures. The identity and purity of HSC were readily confirmed by UV excited autofluorescence of intracellular retinoids. By this standard, more than 95% of cells were found to be vitamin A-positive. After a 2-day culture in the collagen lattice, the cells were immunostained by the anti-rat MMP-9 monoclonal antibody. As shown in Fig. 1E, MMP-9 immunoreactivity was present only in the cells stimulated with IL-1α. We also confirmed the type IV collagenase activity as MMP-9 by a Western blot analysis of the 6-day cultured medium (Fig. 1F).

These results indicate that IL-1α-induced HSC activation in three-dimensional collagen is tightly associated with cytokine-provoked expression and proteolytic activation of multiple ECM proteinases. They also suggest that the breakdown of the ECM which helped maintain the quiescent state of HSC is essential for IL-1α-induced activation of HSC. Conversely, in the absence of ECM framework the HSC grown on plastic readily undergo activation without the need of MMPs.

Specific ECM and Cytokines Are Both Required for Pro-MMP-9 Expression and Activation by HSC

We addressed next the specificity of ECM required for the IL-1α-mediated MMP-9 induction and activation. We tested Matrigel, which contains the basement membrane components similar to that found in the space of Disse between HSC and sinusoidal endothelial cells. We also tested fibrin clot, which is frequently present in the early stage of acute wounds. These two matrices were compared with fibrillar type I collagen, which is normally present between HSC and hepatocytes and becomes abundant in fibrotic conditions. Quiescent HSC were embedded in the ECM followed by cytokine stimulation. We found very interesting patterns of cytokine-mediated regulation of MMP-9 expression/activation by the HSC in these matrices. In Matrigel, IL-1α and TNF-α but not TGF-β induced expression of pro-MMP-9. However, the 105-kDa pro-MMP-9 remained as a zymogen without being converted to the active form. When HSC were grown in type I collagen, IL-1α induced pro-MMP-9 expression followed by conversion to the active 95-kDa form, whereas TNF-α failed even to induce pro-MMP-9. In fibrin clot HSC did not secrete MMP-9 in response to all of these cytokines. Thus these results highlight that the ECM, presumably through the specific cellular receptors, can drastically determine the responsiveness of HSC for their cytokine-mediated regulation of MMP-9 expression and activation.

IL-1α and Type I Collagen Serve as Costimulatory Signals for Induction of MMP-9 and MMP-13, whereas Collagen Is Not Required for MMP-3 Expression

We evaluated cytokines and ECM on other MMPs represented in fibrosis. As shown in Fig. 3A, we confirmed that dual signals from IL-1α and type I collagen are required for induction and activation of MMP-9. After a prolonged development of the zymogram, we occasionally observed weak induction of pro-MMP-9 by the IL-1α-treated HSC on plastic.

We also examined cytokine-mediated regulation of MMP-3 and MMP-13 in this model. MMP-1, an interstitial collagenase (collagenase-1) is absent in the rat and mouse genome, whereas MMP-13 (collagenase-3) is regarded as its counterpart in these species (46). Very similar to the mode of MMP-9 induction, dual signals from IL-1α and type I collagen are required to induce MMP-13 by HSC (Fig. 3B). A 60-kDa MMP-13 and a fragment of ~25 kDa were recognized by a monoclonal antibody for the MMP. TNF-α, TGF-β, and PDGF individually or in combination with collagen all failed to induce MMP-13. This IL-1α-specific induction of MMP-13 paralleled the breakdown of the fibrillar collagen lattices (Fig. 3C). Conversely, the absence of MMP-13 in response to the stimulation by TNF-α, TGF-β, and PDGF is consistent with their incompetence for induction of collagen breakdown. Thus the global meltdown of fibrillar collagen gel is likely attributable to the participation of MMP-13.

MMP-3 (stromelysin-1) is a wide spectrum proteinase for many ECM components. As shown in Fig. 3D, of the panel of cytokines, only IL-1α induced secretion of MMP-3. In contrast to MMP-9 and MMP-13 induction, collagen signal seemed not required for the MMP-3 induction. Finally, MMP-2 was constitutively expressed by HSC, and these cytokines seemed without effect on its expression or activation. Thus, these differential requirements of ECM for IL-1α-mediated induction of MMPs may be of pathophysiological importance because the ECM environment surrounding HSC dynamically changes in liver fibrogenesis, and this may in turn determine the responsiveness of HSC in MMP expression. These results also suggest that similar intracellular signaling may be involved in the regulation of MMP-9 and MMP-13 expression, whereas MMP-3 induction is collagen-independent.

Inactivation of MMP-9 by TIMP-1 or Depletion of MMP-9 by Gelatin-Sepharose 4B Prevents IL-1α-induced HSC Activation in Three-dimensional Collagen

Pro-MMP-9 forms a complex with TIMP-1, which inhibits both the catalytic activity and the zymogen activation (47). Because of the association of HSC activation with pro-MMP-9 expression/activation in IL-1-treated HSC in three-dimensional collagen, we assessed the causal role of MMP-9 in HSC activation with the use of TIMP-1. Primary HSC were embedded in three-dimensional collagen or plated on plastic as a monolayer followed by treatment with IL-1α in the presence or absence of purified recombinant TIMP-1. At the basal state (control), the perinuclear fat droplets were evident in HSC grown in collagen matrix even after culture for 5 days (difficult to observe at this magnification, Fig. 4A). On the treatment of IL-1α, the expression of α-SMA became prominent, and fat droplets were reduced markedly in HSC in three-dimensional collagen and more so in those on plastic. TIMP-1 at 10 nM decreased expression of α-SMA in IL-1α-treated HSC cultured in collagen but not in those on plastic. With a higher concentration of TIMP-1 (30 nM), IL-1-induced expression of α-SMA was conspicuously inhibited. HSC cytoplasm was retracted, and the morphological features of quiescent HSC were restored, including more abundant lipid droplets (Fig. 4A). In contrast, TIMP-1 even at this high concentration had no effect on the IL-1α-induced HSC activation on plastic. This is important because it indicates that the TIMP-1-mediated inhibition of HSC activation in collagen gel is not likely mediated via its direct effects on the cells, but rather through inhibition of extracellular proteinase(s). Indeed, the TIMP-1 treatment inhibited the type IV collagenase activities in the conditioned medium as measured by the fluorogenic peptide cleavage (Fig. 4B). At 30 nM TIMP-1 also prevented IL-1α-induced conversion of the 105-kDa pro-MMP-9 into the active enzyme (Fig. 4C). Finally, TIMP-1 also blocked the IL-1α-mediated breakdown of collagen lattice (Fig. 4D).

Although TIMP-1 shows preferential activity toward MMP-9, it may also target other MMPs. To define the unique role of MMP-9 in HSC activation we depleted the proteinase by using gelatin-conjugated Sepharose 4B in the culture. For this, HSC were embedded into three-dimensional type I collagen in the bottom chamber of Transwell and treated with IL-1α. Gelatin-conjugated Sepharose 4B or control beads were loaded into the upper insert chambers. After a 6-day culture, the HSC with gelatin beads retained the quiescent morphology, whereas the cells with the control beads underwent activation (Fig. 5A). MMP-9 activity in the conditioned medium was reduced significantly by the use of gelatin beads but not by the control beads (Fig. 5B). Consistently, fibrillar collagen lattice was mostly intact in the culture with gelatin beads but was degraded in the control culture with carrier beads (Fig. 5C). Because MMP-2 accounts for a very small portion of the total gelatinase activities, the suppressed activation of HSC with gelatin beads is due mostly to the deprivation of MMP-9. Taken together, the results unequivocally demonstrate that the MMP-9 activity is critical for IL-1α-induced activation of HSC presumably through a dynamic remodeling of ECM. Based on these results we conclude that MMP-9 plays a pivotal role in HSC activation in this culture model.

Fully Activated HSC Reduced MMP Production

A critical biochemical feature of the progressive liver fibrosis is a relative decline in proteinase activities favoring increased deposition of collagen matrix (27). Thus we questioned how fully activated HSC would behave in response to IL-1α in three-dimensional type I collagen. To generate fully activated HSC, the primary HSC were cultured on plastic with 10% FBS for 10 days. By then, the cells were fully trans-differentiated into myofibroblastic cells as judged by the formation of prominent stress fibers and depletion of lipid droplets. At this point, the cells were reseeded on plastic or embedded in three-dimensional collagen lattice followed by cytokine stimulation. Unlike the quiescent HSC, no MMP-9 was secreted by these fully activated cells on plastic or in three-dimensional collagen either basally or in response to cytokine (Fig. 6A). Further, collagen gel was not degraded by these activated HSC under the cytokine stimulation.

We also measured production of MMP-13 and MMP-3 by the fully activated HSC. As shown in Fig. 6B, no MMP-13 was detected by myofibroblastic cells grown on plastic in response to the cytokine stimulation. A minor amount MMP-13 was produced by the fully activated cells in the collagen lattice in response to IL-1α. The fully activated HSC on plastic constitutively secreted a moderate amount of MMP-3, whereas this expression disappeared when embedded in collagen (Fig. 6C). All of these results from the myofibroblastic HSC agree well with the clinical observed down-regulation of MMPs in fibrotic liver.

Colocalization of MMP-9 with α-SMA in a Subpopulation of HSC in Human Fibrotic Liver Tissues

We wanted to verify whether MMP-9 expression could be localized in HSC in liver fibrosis. To this end, we performed double immunostaining for MMP-9 and α-SMA in fibrotic liver biopsy specimens from two patients. As shown in Fig. 7, MMP-9 and α-SMA immunostaining were colocalized in clusters of the fibrotic regions. Although MMP-9 and α-SMA were largely costained in these two patients, their patterns are slightly different. In one patient most of the MMP-9-positive cells were colocalized with α-SMA staining (Fig. 7A). In the other patient, some MMP-9-positive cells were not stained for α-SMA (Fig. 7B). We speculate that these double stained cells may represent a subpopulation of HSC in the process of activation like rat HSC preceding activation in response to IL-1 in three-dimensional collagen. The MMP-9-negative/α-SMA-positive cells may be the fully activated HSC which expressed abundant α-SMA but no MMP-9. The MMP-9-positive/α-SMA-negative cells are likely to be macrophages (48) or hepatocytes (49).