Platelet-derived Growth Factor-BB and Thrombin Generate Positive and Negative Signals for Human Hepatic Stellate Cell Proliferation

Proliferation of myofibroblastic hepatic stellate cells (HSC) in response to growth factors is essential for the development of liver fibrosis. We have reported that prostaglandins (PG) and cyclic AMP (cAMP) inhibit growth of human HSC. This PG/cAMP pathway transduces the endothelin (ET) B-mediated antiproliferative effect of endothelin-1 (ET-1) and up-regulates ETB receptors. Here, we show that platelet-derived growth factor (PDGF)-BB and thrombin, although mitogenic, generate growth inhibitory PGE2in myofibroblastic human HSC. The two peptides elicit early PGE2 and cAMP synthesis, and also promote delayed induction of cyclooxygenase (COX)-2. Both early and delayed production of PGE2 counteract the mitogenic effect of PDGF-BB and thrombin because: (i) pretreatment with the COX inhibitor ibuprofen markedly enhances the mitogenic effect of both peptides; (ii) blocking early synthesis of PGE2 greatly enhances extracellular signal-regulated kinase (ERK) activation by both growth factors; (iii) enhancement of DNA synthesis by ibuprofen is only lost when the inhibitor is added after COX-2 induction has occurred. Finally, PDGF-BB and thrombin raise ETB receptors through the PG pathway. Thus, ibuprofen blunts growth factor-induced increase in ETB receptors. Up-regulation of the growth inhibitory ETB receptors by both mitogens may enhance the antiproliferative effect of ET-1 and thereby establish a negative feedback of their mitogenic effect. Our results shed light on novel growth inhibitory signals evoked by two mitogenic growth factors expressed during liver injury.

. In response to growth factors and cytokines expressed by neighboring cells, myofibroblastic HSC display enhanced mitogenicity and synthesize increased amounts of extracellular matrix (1). Experimental models of liver injury have shown that increased proliferation of myofibroblastic HSC results in accumulation of cells in the liver (3,4). Among soluble factors that enhance HSC proliferation, PDGF-BB is the most potent mitogen for myofibroblastic HSC in culture and its effects have been extensively studied (5,6). Expression of PDGF and of its tyrosine kinase receptors is induced during the process of activation, as shown in cirrhotic human liver, in experimental models of liver injury, and in cultured HSC (7)(8)(9). Thrombin is also mitogenic for these cells, and myofibroblastic HSC show increased expression of its receptor during liver injury (10,11).
Given the highly proliferative capacity of myofibroblastic HSC, several studies have focused on factors that may limit their growth (12)(13)(14)(15). We and others have recently shown that PGE 2 and cAMP inhibit proliferation of human myofibroblastic HSC (15,16). We also found that binding of endothelin-1 (ET-1) to its ETB receptor stimulates production of prostaglandins that in turn elevate cAMP, resulting in a growth inhibitory effect of the peptide (13,15). In addition, PG and cAMP upregulate ETB receptors, which may amplify the antiproliferative effect of ET-1 (13,15).
Although PDGF-BB stimulates cell proliferation by activating key effectors, including ras, raf, and the extracellular signal-regulated kinase (ERK) cascade (17), it has also been reported that the peptide may increase synthesis of cAMP (18 -22). Thrombin inhibits adenylyl cyclase in most cells, but increases in cAMP production have also been reported in a few instances (23)(24)(25). In the present study, we have used a model of cultured human myofibroblastic HSC that displays the phenotypic and functional characteristics of HSC found in situ during hepatic fibrogenesis. (26). We show that, in myofibroblastic HSC, thrombin and PDGF-BB evoke a growth inhibitory PG/cAMP signal that reduces their promitogenic effects by (i) counteracting the transduction pathway stimulated by mitogenic signals and (ii) increasing the density of growth inhibitory ETB receptors.
Cell Isolation and Culture-Human HSC were obtained in their activated phenotype by outgrowth from explants of normal liver ob-tained following surgery of benign or malignant liver tumors. This procedure is in accordance with ethical regulations imposed by the French legislation. Explants were incubated, as described previously, in Dulbecco's modified Eagle's medium containing 10% serum (5% fetal calf serum, 5% pooled human serum) (26). Exhaustive characterization of these cells has already been published (26). Cell isolates were routinely characterized by a positive staining for smooth muscle ␣-actin, a marker of HSC in their myofibroblastic phenotype. Experiments were performed between passages 3 and 8 without any noticeable difference in results observed with cells obtained from various passages, or from various livers. All experiments were performed on confluent cells made quiescent in serum-free Waymouth medium over 3 days.
cAMP Assay-Cells were seeded in 24-well plates, grown to confluence, and deprived of serum for 48 h in Waymouth medium. Following preincubation with 0.6 mM isobutylmethylxanthine for 15 min, HSC were stimulated in phosphate-buffered saline with effectors. When indicated, cells were preincubated for 30 min with the cyclooxygenase inhibitor ibuprofen (50 M) or vehicle (ethanol). Assays were performed in duplicate. cAMP was extracted in 95% ethanol for 2 h at 25°C, and extracts were stored at Ϫ80°C, pending assay (15). cAMP was assayed by a commercial radioimmunoassay (Immunotech, France).
Prostaglandin Release-Confluent monolayers in 24-well plates were deprived of serum for 48 h, washed twice in phosphate-buffered saline (PBS), and further incubated in Waymouth medium. Following pretreatment with ibuprofen 50 M or vehicle, cells were stimulated with growth factors as indicated, over various periods of time. Release of PGE 2 was assayed by a specific enzyme immunoassay, as described previously (27).
DNA Synthesis Assay-DNA synthesis was measured in triplicate wells by incorporation of [ 3 H]thymidine, as described previously (14). Confluent HSC were made quiescent by incubation in Waymouth medium without serum for 48 h. Cells were then stimulated for 30 h with growth factors, following pretreatment for 30 min with ibuprofen (50 M) or vehicle. [ 3 H]Thymidine (0.5 Ci/well) was added during the last 6 h of incubation.
Thymidine incorporation was also assessed by nuclear autoradiography (12). Briefly, confluent quiescent HSC in 35-mm plates were stimulated for 30 h with growth factors following treatment with ibuprofen or its vehicle. [ 3 H]Thymidine (2 Ci/ml) was added during the last 6 h of incubation. Cultures were washed in PBS, fixed in 3% paraformaldehyde for 30 min, and dehydrated with 100% methanol. They were covered with LM1 emulsion (Amersham, France) and exposed for 3 days. Labeling was developed with Kodak Dektol developer and fixed with Kodak Unifix. Cells were counterstained in eosin and hemalun. Results were expressed as the ratio of nuclei with overlapping silver grains to the total number of nuclei. Data from 5 to 10 random low magnification fields per culture were averaged.
ERK Assays-Confluent cells in 60-mm wells were deprived of serum for 48 h and further stimulated with growth factors following preincubation with ibuprofen (50 M, 30 min) or vehicle (ethanol). The reaction was terminated by washing the cell cultures twice with 6 ml of icechilled PBS. Cells were lysed for 30 min at 4°C in 0.3 ml of 50 mM HEPES, pH 7.4, containing 1% Triton X-100, 10% glycerol, 137 mM NaCl, 1.5 mM MgCl2, 10 mM NaF, 0.1 mM dithiothreitol, 1 mM EGTA, 5 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin, 50 g/ml pepstatin, 40 mM ␤-glycerophosphate, and 10 mM paranitrophenylphosphate. Lysates were centrifuged at 4°C for 15 min at 12,000 ϫ g. The supernatants were either used directly, or frozen at Ϫ70°C. Extracellular signal-regulated kinase activity was assayed in situ following electrophoresis of equal amounts of cell lysates (40 g of proteins), on a 10% SDS-polyacrylamide gel copolymerized with 0.5 mg/ml myelin basic protein, as described previously (13). Phosphorylated gels were exposed to x-ray films, and ERK activity was quantified by using a PhosphorImager (Molecular Dynamics, France).
After electrophoresis on a denaturing SDS gel, proteins (20 g/lane) were electroblotted onto nitrocellulose membranes and blocked for 1 h at room temperature in 10 mM Tris, pH 8, containing 150 mM NaCl, 0.05% Tween 20, 5% skimmed milk. Detection of COX-1 (28) and COX-2 (29) was performed after incubation for 2 h at room temperature with their specific antibody diluted 1:2000-and 1:5000-fold, respectively. After 1 h of incubation with 2000-fold diluted horseradish peroxidaseconjugated anti-IgG antibody and extensive washings, immunodetected proteins were visualized by using an enhanced chemiluminescence (ECL) assay kit (Amersham) according to the instructions of the manufacturer.
Binding of [ 125 I]-ET-1 to a Particulate Fraction of HSC-Confluent monolayers were deprived of serum for 48 h and subsequently stimulated with growth factors, following pretreatment with 50 M ibuprofen or vehicle (ethanol). A particulate fraction was obtained (13) and binding of 125 I-ET-1 was measured as described previously (30). Briefly, the HSC particulate fraction (5-10 g of proteins) was incubated for 90 min at 22°C in 200 l of Krebs-Ringer medium containing 20 mM HEPES, pH 7.4, 1% (w/v) bovine serum albumin, 300 g/ml bacitracin, with either 50 pM 125 I-ET-1 and varying concentrations of cold ET-1, sarafotoxin S6C or BQ123 (competition experiments), or varying concentrations of 125 I-ET-1 (4.5-800 pM, saturation experiments). Nonspecific binding was determined by incubating with 0.1 M unlabeled ET-1. Data from saturation and competition experiments were analyzed using the nonlinear regression program LIGAND.
Assay of Protein Concentration-Protein concentration was measured using a commercial reagent (Bio-Rad, France) Statistics-Results were analyzed by 2-way analysis of variance (ANOVA) for repeated measurements, Wilkoxon's signed rank test, or unpaired Student's t test, as appropriate.

PDGF-BB and Thrombin Stimulate the Synthesis of the Growth Inhibitory Messengers PGE 2 and cAMP in Human
Myofibroblastic HSC-Treatment of human myofibroblastic HSC with mitogenic concentrations of thrombin and PDGF-BB increased the production of PGE 2 (Fig. 1A). Prostaglandin E 2 synthesis increased within 2 min in the presence of thrombin and was maximally elevated by 4-fold over basal after 5 min. The time course of the effect of PDGF-BB on PGE 2 release was slightly delayed, rising after 5 min and reaching a maximal 4-fold increase over basal at 30 min. The COX inhibitor ibuprofen blunted the elevation of PGE 2 production elicited by PDGF-BB and thrombin. Both PDGF-BB and thrombin also stimulated cAMP production (Fig. 1B). In the presence of thrombin, cAMP concentration rose within 1 min, maximally increased to 7-10-fold by 2 min, and rapidly declined after 10 min. Following PDGF-BB treatment, cAMP increased by 5-fold within 10 min and decreased gradually thereafter. Thrombin and PDGF-BB increased cAMP through prostaglandin production because ibuprofen blunted the enhancement of cAMP production elicited by both peptides (Fig. 1C). Similar results were obtained when cells were stimulated with 5% human serum instead of purified growth factors (not shown). In contrast, IGF-1 at a concentration mitogenic for HSC (100 ng/ml) (Fig.  3), did not affect either PGE 2 or cAMP levels (Fig. 1, A and B). Ibuprofen also decreased basal cAMP levels from 330 to 60 fmol/mg protein by blocking endogenous synthesis of PG (Fig. 1C).
Taken together, these results indicate that PDGF-BB and thrombin, but not IGF-1, stimulate synthesis and secretion of PGE 2 (Fig. 1), a prostaglandin which increases cAMP production in an autocrine fashion (15). Inasmuch as PGE 2 and cAMP inhibit growth factor-stimulated proliferation of myofibroblastic HSC (Fig. 1B, inset; Ref. 15), our results indicate that both PDGF-BB and thrombin, although mitogenic for HSC, generate second messengers that are growth inhibitory for these cells.
Stimulation of the PG/cAMP Pathway by PDGF-BB and Thrombin Down-regulates Their Mitogenic Effects-The following experiments were undertaken to assess the consequences of the stimulation of the PG/cAMP pathway on the mitogenic properties of PDGF-BB and thrombin. Therefore, DNA synthe-sis and ERK activation were assayed following blockade of this pathway by the COX inhibitor ibuprofen.
It is well established that activation of ERK by growth factors is an important early step preceding cell proliferation (17,31). PDGF-BB rapidly stimulated ERK, inducing a maximal activation within 5 to 10 min, followed by a decrease to a sustained level for at least 1 h ( Fig. 2A). Stimulation of ERK by thrombin showed a peak at 5-10 min and returned to basal levels within 1 h (Fig. 2B). Pretreatment of cells with ibuprofen pretreatment markedly potentiated the activation profile of ERK by PDGF-BB or thrombin. In the presence of the COX inhibitor, peak activation of ERK by PDGF-BB was achieved after 10 min. The compound did not affect stimulation of ERK by PDGF-BB at early time points but synergistically enhanced ERK activation from 10 min onward ( Fig. 2A). Activation of ERK by thrombin was also markedly potentiated by ibuprofen, although as early as 5 min (Fig. 2B). Finally, the COX inhibitor moderately enhanced basal ERK activity, by preventing endogenous release of PG. It should be noted that PDGF-BB has a delayed effect on PGE 2 and cAMP synthesis, as compared with thrombin ( Fig. 1). This time frame is in keeping with the potentiating effect of ibuprofen on ERK activation by PDGF-BB and thrombin (Fig. 2).
Ibuprofen pretreatment synergistically enhanced the mitogenic effects of thrombin and PDGF-BB at all concentrations tested (Fig. 3, A and B). The COX inhibitor increased basal DNA synthesis by 2-fold. In contrast, ibuprofen pretreatment maximally increased DNA synthesis of PDGF-BB-treated cells by 10 -11-fold, as compared with the 4-fold increase in cells exposed to PDGF-BB alone (p Ͻ 0.001 by 2-way ANOVA). Thrombin alone caused a 2-fold increase of thymidine incorporation, which rose to 7-fold following ibuprofen pretreatment (p Ͻ 0.001 by 2-way ANOVA). Similar results were obtained when cells were stimulated with 5% human serum instead of purified growth factors (not shown). As expected, the cyclooxygenase inhibitor did not potentiate the effect of IGF-1 on DNA synthesis (Fig. 3C). The synergistic effect of ibuprofen on PDGF-BB or thrombin-induced mitogenesis was confirmed by nuclear autoradiography experiments (Fig. 3D). Altogether, these data indicate that in myofibroblastic HSC, PDGF-BB and thrombin stimulate a PG/cAMP pathway that reduces activation of ERK and decreases their effects on cell proliferation.
The role of the PG/cAMP pathway was further supported by experiments in which ibuprofen was added at various times before or after the introduction of growth factors. As shown above, introduction of ibuprofen 30 min before the addition of thrombin (Ϫ30 min) maximally enhanced the mitogenic effect of the peptide (Fig. 4 A). Addition of ibuprofen 30 min after that of thrombin (ϩ30 min) also synergistically enhanced its mitogenic effect although to a lower extent (5-fold at ϩ30 min, as compared with 9.5-fold at Ϫ30 min, p Ͻ 0.001 by 2-way ANOVA). The potentiating effect of ibuprofen was no longer observed when the compound was added 6 h after thrombin. Similar results were obtained when cells were stimulated with PDGF-BB instead of thrombin (Fig. 4B). These data indicate that growth factors stimulate early as well as late production of PGE 2 .
Late synthesis of PGE 2 could be demonstrated by studying the regulation of inducible type 2 cyclooxygenase (COX-2) expression. As shown in Fig. 5, a 6-h treatment with PDGF-BB and thrombin up-regulated COX-2 expression, whereas IGF-1 had no effect. In contrast, PDGF-BB and thrombin did not alter COX-1 protein expression (Fig. 5).
Altogether, these results show that both early and late pro- duction of prostaglandins account for the negative regulation of the mitogenic effect of PDGF-BB and thrombin.

PDGF-BB and Thrombin Up-regulate ET Receptors via a Prostaglandin/cAMP Pathway in Human
Myofibroblastic HSC-We have previously shown, that binding of ET-1 to its ETB receptors inhibits proliferation of human HSC, and that exogenously applied PGE 2 and/or cAMP up-regulate ETB receptors ( Refs. 13 and 15; Fig. 7). Therefore, we investigated whether PDGF-BB and thrombin may increase the growth inhibitory ETB receptors via the PG/cAMP pathway.
Treatment of HSC for 24 h with PDGF-BB (20 ng/ml) or thrombin (5 units/ml) increased ET-1 binding sites by 2.2-and 1.5-fold, respectively, whereas IGF-1 had no effect (Fig. 6). There was no significant change in the K d with either peptide (not shown). The increase in ET receptors was time-dependent. In PDGF-BB-treated cells, the number of receptors rose by 2.2-fold within 14 h and remained elevated until 24 h (Fig. 6A). In thrombin-treated cells, a maximal 1.5-fold increase was found after 3 h, which was unchanged over 24 h (Fig. 6B). We also assessed the effect of PDGF-BB and thrombin on the distribution of ETA and ETB receptors. In control cells, ETA and ETB receptors were present in a proportion of 55:45. Growth factors did not modify the proportion of ET receptors (50:50), so that both ETA and ETB receptors were up-regulated by 2.2-and 1.5-fold in PDGF-BB and thrombin-treated cells, respectively (Fig. 7).
We next examined the role of the PG pathway in the regulation of ET receptors. In unstimulated cells, ibuprofen alone did not affect the number of ETA receptors (control: 211 fmol/mg protein; ibuprofen: 191 fmol/mg protein) but selectively down-regulated ETB receptors by 50% (from 180 fmol/mg protein in control cells to 90 fmol/mg protein in ibuprofenpretreated cells) by preventing endogenous PG release. Upregulation of ETA receptors by PDGF-BB and thrombin was not modified following ibuprofen pretreatment (2.0-and 1.5fold, respectively) (Fig. 7). In contrast, ibuprofen totally downregulated ETB receptors in PDGF-BB and thrombin-treated cells (Fig. 7). Collectively, these results indicate that PDGF-BB and thrombin increase both ETA and ETB receptors in myofibroblastic HSC. However, blockade of the PG/cAMP pathway specifically blunts the up-regulation of ETB receptors by both peptides. DISCUSSION In the present study, we demonstrate that in human myofibroblastic HSC, PDGF-BB, and thrombin generate growth inhibitory signals by stimulating a PG/cAMP pathway. Activation of this signal counteracts their mitogenic effects by blocking the transduction pathway leading to HSC prolifera- tion and by up-regulating ETB receptors for the growth inhibitory peptide ET-1 (13).
Binding of PDGF-BB to its tyrosine kinase receptor and of thrombin to its G protein-coupled receptor stimulate proliferation of a number of cells, including myofibroblastic HSC (17,23). A striking point of the present study is that, although mitogenic, PDGF-BB and thrombin also generate growth inhibitory PGE 2 in human myofibroblastic HSC (15). The two peptides elicit early PGE 2 production by activating basal COX, resulting in cAMP accumulation (Fig. 1), and also promote delayed induction of COX-2. Both early and delayed production of PGE 2 negatively regulate the mitogenic effect of PDGF-BB and thrombin. The role of the early synthesis of PGE 2 is demonstrated by two lines of evidence: (i) addition of ibuprofen after the occurrence of the early release of PGE 2 (ϩ30 min, see Fig. 4) reduces the potentiating effect of the COX inhibitor on DNA synthesis; and (ii) ibuprofen pretreatment enhances ERK activation by both growth factors within minutes (Fig. 2). The growth inhibitory signal evoked by PDGF-BB and thrombin also involves delayed induction of COX-2 expression; thus, the potentiating effect of ibuprofen on DNA synthesis is suppressed only when the COX inhibitor is added after induction of COX-2 by both peptides has occurred (Fig. 4). In agreement with these data, we recently observed that induction of COX-2 is involved in the antiproliferative effect of ET-1 and TNF-␣ in human myofibroblastic HSC. 2 We conclude that the mitogenic effects of PDGF-BB and thrombin result from simultaneous stimulatory and inhibitory signals: a growth enhancing pathway, involving activation of ERK; and a growth inhibitory signal mediated by a PG/cAMP pathway, which reduces the mitogenic effect of both growth factors. It should be noted that activation of the PG/cAMP pathway is not a general feature of tyrosine kinase receptor activation. Thus, IGF-1 does not enhance early production of PGE 2 and cAMP, either in human HSC (Fig. 1) or in human vascular smooth muscle cells (18); neither does IGF-1 affect COX-2 protein expression (Fig. 5).
In keeping with our results, recent studies provide evidence that growth factors may positively and negatively regulate cell proliferation, via PG and/or cAMP. In COS-7 cells, isoproterenol raises cAMP and ERK activity, and activation of ERK by isoproterenol is enhanced when blocking the cAMP pathway (32). In human vascular smooth muscle cells, PDGF-BB is mitogenic but also stimulates synthesis of growth inhibitory PG (18,33). Finally, in human fetal lung fibroblasts, indomethacin supresses the antiproliferative effect of TGF␤ and restores its mitogenic effect (34). These observations, together with the results of the present study, strongly suggest that a PG and/or cAMP pathway negatively regulates the mitogenic effects of certain growth factors. Interestingly, in NIH3T3 fibroblasts, PDGF-BB activates SIRP␣1 proteins, which in turn counteract the mitogenic effect of the peptide (35). Whether PG and cAMP may activate SIRP␣1 proteins in human HSC remains to be determined.
We have previously shown that ET-1 inhibits proliferation of human myofibroblastic HSC via its ETB receptors. The present study demonstrates that the PG pathway activated by PDGF-BB and thrombin up-regulates ETB receptors. Thus, both growth factors raise the number of ETB binding sites, and ibuprofen blunts this increase. These results are in line with our previous finding that exogenously added PGE 2 and PGI 2 up-regulate ETB receptors in these cells (15). In contrast, IGF-1 does not affect COX-2 protein expression and prostaglandin production and has no effect on the number of ETB receptors. These results clearly establish that PG production plays a central role in the regulation of ETB receptors in human myofibroblastic HSC. Up-regulation of growth inhibitory receptors by mitogens is an important step in the negative control of cell proliferation (36). For instance, epidermal growth factor up-regulates somatostatin receptors and enhances the antiproliferative of somatostatin in pancreatic cancer cells (36). Also, transforming growth factor ␤ inhibits proliferation of rat HSC and decreases PDGF receptors (37). Therefore, it may be suggested that up-regulation of ETB receptors by PDGF-BB and thrombin enhances the antiproliferative effect of ET-1 and thereby establishes a negative feedback of their promitogenic effect. Along this line, it should be noted that PDGF-BB and thrombin stimulate secretion of ET-1 in a number of cells, including human HSC, and that expression of ET-1 is markedly increased in the liver during chronic liver diseases (38).
Besides ETB receptors, human myofibroblastic HSC also express ETA receptors, which mediate the contractile effect of ET-1 (13,39). PDGF-BB and thrombin increase ETA binding sites by a mechanism unrelated to PG production. Interestingly, thrombin enhances contractility of myofibroblastic HSC (40). In this respect, besides its direct contractile effect, the peptide may also indirectly increase HSC contraction by enhancing the effect of ET-1 following up-regulation of ETA receptors.
In experimental models of liver injury and in human chronic liver disease, myofibroblastic HSC proliferate and accumulate in response to cytokines and growth factors produced locally (1). Accordingly, PDGF expression is increased in infiltrating inflammatory cells and within fibrous septa and several lines of evidence suggest that thrombin is also generated in response to liver injury (8 -10). Moreover, myofibroblastic HSC show increased levels of PDGF and thrombin receptors during liver injury, and expression of the PDGF receptor has been correlated to the activity of the fibroproliferative process (8 -10). Several studies have focused on the molecular mechanisms responsible for the growth-enhancing effects of PDGF-BB and thrombin in human myofibroblastic HSC. Our findings shed light on novel growth inhibitory pathways evoked by these peptides. We show that, in fact, the mitogenic effect of PDGF-BB and thrombin arises from the combined effect of positive and negative signals, the latter transduced by PG and cAMP. The promitogenic signal is counteracted by two PG-dependent mechanisms: (i) a direct negative pathway that inhibits ERK activation and more delayed events involved in proliferation; and (ii) an indirect feedback that may amplify the antiproliferative effect of ET-1 by up-regulating ETB receptors. Our results may have implications for the comprehension of the physiopathology of liver fibrosis. They suggest that the final proliferative response depends on the balance of conflicting positive and negative signals elicited simultaneously by a single growth factor. Further studies are needed to better understand factors that may shift the balance toward growth inhibition and thereby limit proliferation of myofibroblastic HSC during chronic liver disease.