Role of NF-κB in the Antiproliferative Effect of Endothelin-1 and Tumor Necrosis Factor-α in Human Hepatic Stellate Cells

During chronic liver diseases, hepatic stellate cells (HSC) acquire an activated myofibroblast-like phenotype and proliferate and synthesize fibrosis components. Endothelin-1 (ET-1), which inhibited the growth of human myofibroblastic HSC, increased the formation of two NF-κB DNA binding complexes; this effect was also observed with tumor necrosis factor-α (TNF-α). The complexes were identified as the p50/p50 and p50/p65 NF-κB dimers. Activation of NF-κB was associated with the degradation of the inhibitory protein IκB-α; no IκB-β was detected. Activation of NF-κB and degradation of IκB-α were prevented by the NF-κB inhibitors sodium salicylate and MG-132. In addition to cyclooxygenase-1 (COX-1), COX-2 is also constitutively expressed in human HSC, and the use of dexamethasone and of SC-58125, a selective COX-2 inhibitor, revealed that COX-2 accounts for basal COX activity. Moreover, COX-2 mRNA and protein were up-regulated by ET-1 and TNF-α, whereas COX-1 was unaffected. Induction of COX-2 and stimulation of COX activity by ET-1 and TNF-α were prevented by sodium salicylate and MG-132, suggesting that activation of NF-κB by either factor is needed for stimulation of COX-2. Finally, SC-58125 and dexamethasone reduced the growth inhibitory effect of ET-1 and TNF-α, indicating that activation of COX-2 is required for inhibition of HSC proliferation. Taken together, our results suggest that NF-κB, by inducing COX-2 expression, may play an important role in the negative regulation of human myofibroblastic HSC proliferation.

Hepatic stellate cells (1) (HSC, 1 also known as lipocytes, fat-storing cells, or perisinusoidal cells) play a crucial role in the pathogenesis of liver inflammation and fibrosis (1). Following acute or chronic liver injury, HSC transdifferentiate from a quiescent cell containing large retinoid droplets to an activated myofibroblast-like cell (2). In response to growth factors and cytokines expressed by neighboring cells, activated HSC display enhanced mitogenicity and secrete multiple inflammatory cytokines and components of fibrosis (1). Experimental models of liver injury have shown an increased proliferation of activated HSC (3,4), and several studies have focused on factors that may limit growth of these cells. In particular, we have shown that binding of endothelin-1 (ET-1) to its G-proteincoupled ETB receptor leads to inhibition of human activated myofibroblastic HSC proliferation, following prostaglandin production (5,6).
NF-B is an essential regulator of the expression of viral genes, as well as of a number of genes involved in immune, inflammatory, and growth responses (7)(8)(9). This family of transcription factors comprise at least five members, p65 (RelA), p50, p52, RelB, and the product of the c-Rel protooncogene, which form homo-or heterodimers and bind to a DNA sequence called the B motif. In most cell types, NF-B is present in the cytosol in an inactive form, as a heterodimer of a 50-kDa (p50) and 65-kDa (p65) subunit, bound to one of the IB inhibitory proteins. Of the different IBs, IB-␣ and the more recently identified IB-␤ have been well characterized. All known activators of NF-B (cytokines, phorbol esters, growth factors, and viral trans-activators) rapidly phosphorylate IB-␣, resulting in its ubiquitination and degradation through a proteasomedependent pathway. Degradation of IB-␤ occurs more slowly and is induced by agonists that cause persistent activation of NF-B (10). The free NF-B dimer subsequently translocates into the nucleus, where it activates gene transcription (7)(8)(9)11). Resynthesis of IB-␣ is induced by NF-B and allows resequestration of NF-B in the cytoplasm, shutting down the NF-B response (11).
Cyclooxygenases (COX) are the rate-limiting enzymes in the biosynthetic pathway of prostaglandins and thromboxanes from arachidonic acid. These enzymes catalyze the conversion of arachidonic acid to PGH2, the committed step in prostanoid synthesis. Two isozymes are found in mammalian tissues, COX-1 and COX-2. The COX-1 isoform is constitutively expressed in most cells, and the COX-2 gene, generally absent in resting cells, is rapidly induced by hormones, cytokines, and tumor promoters (12). Sequencing of the human cyclooxygenase-2 promoter region has revealed the presence of two NF-B consensus sites, which are important in the induction of the COX-2 mRNA by TNF-␣ (13).
Since some evidence supports a role of NF-B in the activation of genes that regulate cell growth arrest (14 -16), in this study we investigated whether activation of NF-B may be involved in the negative control of human myofibroblastic HSC proliferation. We show that two factors that inhibit the growth of human HSC, ET-1 (5, 6) and TNF-␣ (the present study), activate NF-B in these cells. We also report that activation of NF-B by both factors leads to the induction of cyclooxygenase-2 (COX-2), which mediates inhibition of human HSC growth. All these results suggest that NF-B, by inducing COX-2 expression, may play an important role in the negative regulation of human myofibroblastic HSC proliferation.
Isolation and Culture of Human Myofibroblastic HSC-Human HSC in their activated myofibroblastic phenotype were obtained by outgrowth of normal liver explants obtained from surgery of benign or malignant liver tumors. This procedure was performed in accordance with ethical regulations imposed by the French legislation. Explants were incubated in Dulbecco's modified Eagle's medium containing 10% serum (5% fetal calf serum, 5% pooled human serum), and exhaustive characterization of these cells has already been published (17). 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 7, 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.
DNA Synthesis Assay-DNA synthesis was measured in triplicate wells by incorporation of [ 3 H]thymidine, as described previously (6). Following pretreatment for 60 min with 25 M SC-58125 or 18 h with dexamethasone or vehicle, confluent quiescent cells were stimulated for 30 h with TNF-␣ or ET-1, in the presence of 5% human serum. [ 3 H]Thymidine (0.5 Ci/well) was added during the last 6 h of incubation.
Nuclear and cytoplasmic extracts were prepared as described by Dignam et al. (18) with minor modifications. Confluent cells in 10-cm dishes were made quiescent in serum-free Waymouth medium over 3 days and were further incubated for various times with the indicated effectors. Cells were then washed two times in 10 ml of ice-cold phosphate-buffered saline and resuspended in 400 l of buffer A (10 mM Hepes, pH 7.9, containing 1.5 mM MgCl 2 ,10 mM KCl, 0.5 mM DTT, 0.5 mM PMSF, 1 g/ml leupeptin, 1 g/ml aprotinin, 1 g/ml pepstatin A). The cells were allowed to swell on ice for 15 min, after which 12.5 l of 10% Nonidet P-40 was added. The tubes were shaken gently and centrifuged at 2,000 ϫ g for 10 min at 4°C, and supernatants were used as cytoplasmic extracts. The pellet nuclei were resuspended in 40 l of buffer C (20 mM Hepes, pH 7.9, containing 1.5 mM MgCl 2 , 450 mM NaCl, 25% glycerol, 0.2 mM EDTA, 0.5 mM DTT, 0.5 mM PMSF, 1 g/ml leupeptin, 1 g/ml aprotinin, 1 g/ml pepstatin A). After 30 min at 4°C under constant agitation, nuclear debris were centrifuged at 20,000 ϫ g for 15 min. The supernatants (nuclear extract) were frozen in liquid nitrogen and stored at Ϫ80°C.
Preparation of DNA Probes for Electrophoretic Mobility Shift Assay-Sense and antisense oligomers containing NF-B-binding sites were annealed, and the double-stranded oligomer was radiolabeled with T4 polynucleotide kinase and [␥-32 P]ATP. The unincorporated nucleotides were removed by filtration though a G50 Fine column. Sequences of the oligomers containing the positive regulatory element (PRE) B sequence of the HIV-LTR were 5Ј-ACAAGGGACTTTCCGCTGGGGACTTTCCAGG-3Ј, and oligomers containing the B-responsive element of the COX-2 promoter were 5Ј-GGAGAGGGGATTCCCTGCGC-3Ј. In some experiments, we also used with similar results an oligomer (Promega) corresponding to the consensus sequence of NF-B from the light chain enhancer. The sequence of the oligomer containing the HNF3 sequence of the HNF1 promoter was 5Ј-TCGAGCTCCAATGTAAACAGAGCAGGT-3Ј. Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts (5-10 g of protein) were incubated in the binding reaction medium (20 mM Hepes, pH 7.9, 100 mM KCl, 20% glycerol, 0.2 mM EDTA, 0.5 mM DTT, 0.5 mM PMSF) for 15 min at 4°C, followed by a 15-min incubation with 0.5 ng of 32 P-labeled oligonucleotide at room temperature. The DNA-protein complexes were analyzed on a 5% polyacrylamide gel in 0.25ϫ Tris borate/EDTA electrophoresis buffer. Gels were run at 150 V for 90 min, dried, and autoradiographed. In competition assays, 100 ng of either unlabeled NF-B, HNF3, or COX-2 oligonucleotides were added to the binding reaction for 15 min before addition of the radiolabeled probe. In the supershift experiments, the binding reaction containing the antibodies were incubated for 1 h at room temperature before adding the radioactive probe.
RNA Preparation and Northern Blot Analysis-Confluent cells were made quiescent in serum-free Waymouth medium over 3 days and were then stimulated with 0.1 M ET-1 or 50 ng/ml TNF-␣, as indicated. Total RNA was extracted in guanidinium thiocyanate, as described previously (5). RNA samples (20 g/lane) were denatured, fractionated by electrophoresis through a 1% agarose/formaldehyde gel, and subsequently transferred to a nitrocellulose membrane (Schleicher & Schuell). Prehybridization was performed at 42°C in 5ϫ SSC (1ϫSSC ϭ 150 mM NaCl, 15 mM sodium citrate) containing 50% formamide, 1ϫ Denhardt's, 20 mM NaH 2 PO 4 , 10% dextran, 170 g/ml heparin, and 200 g/ml salmon sperm DNA. cDNA probes for COX-2 and human glyceraldehyde-3-phosphate dehydrogenase (G3PDH) were labeled with [␣-32 P]dATP using a random priming Ready to Go kit (Pharmacia). Blots were hybridized overnight at 42°C with the 32 P-labeled probes, washed at 22°C for 1 min with 2ϫ SSC, 0.1% SDS and once in 0.2ϫ SSC, 0.1% SDS at 55°C and exposed to x-ray film (Hyperfilm, Amersham) at Ϫ70°C. Hybridization signals were quantified by Fuji Bio-imaging analyzer (Fuji, Tokyo, Japan). Results are relative to G3PDH expression, which was used as an internal standard to correct for variations in loading and transfer. Results are expressed as fold over basal COX-2/G3PDH ratio.
Cyclooxygenase Activity and PGE 2 Release-Confluent cells were made quiescent in serum-free Waymouth medium over 3 days and were then pretreated for 60 min with either 25 M SC-58125, 25 M ibuprofen, 20 mM sodium salicylate, 50 M MG-132 or 18 h with 0.1 M dexamethasone. Cells were further stimulated for the time indicated with 0.1 M ET-1 or 50 ng/ml TNF-␣. When indicated, 10 M arachidonic acid, freshly diluted in 0.1% bovine serum albumin, was added for 30 min at the end of incubation. Release of PGE 2 in the medium was assayed as described previously (6).
Assay of Protein Concentration-Protein concentration was determined by Bio-Rad protein assay kit (Bio-Rad, France) according to the manufacturer's instructions.
Statistics-Results are expressed as mean Ϯ S.E. of n experiments. Statistical analysis was determined by Student's unpaired t test with p Ͻ 0.05 considered significant.

RESULTS
ET-1, the ETB Receptor Agonist Sarafotoxin-S6C, and TNF-␣ Activate NF-B in Human Myofibroblastic HSC-Nuclear proteins isolated from ET-1 or TNF-␣-treated human myofibroblastic HSC were analyzed in electrophoretic mobility gel shift assays (EMSA), using a radiolabeled DNA probe containing the B sequence of the HIV-LTR promoter. As shown in Fig. 1A (lane 2), ET-1 increased the formation of two NF-B-DNA binding complexes. This activation was also observed with the ETB receptor agonist sarafotoxin S6C (Fig. 1A, lane  3). TNF-␣, which activates NF-B in numerous cell types (7-9), increased DNA binding complexes with profiles similar to ET-1 (Fig. 1A, lane 4).
The specificity and composition of the complexes stimulated by ET-1 and TNF-␣ were next examined. Competition experiments, performed with an excess of unlabeled nucleotide corresponding either to the HIV-LTR B or to one of the Bresponsive elements of the human COX-2 (13), eliminated the binding activity of the labeled probe (Fig. 1B, lanes 6 and 8). In contrast, addition of the same excess of an unlabeled oligonucleotide containing the unrelated HNF3 consensus sequence had no effect (Fig. 1B, lane 7). These data indicate that the DNA binding activity increased with ET-1, and TNF-␣ is specific for the B motif. Supershift experiments using antibodies to each of the NF-B/Rel proteins were performed to identify the proteins present in the NF-B binding complexes. Incubation of nuclear extracts from ET-1-treated cells with p50 antibodies caused a supershift of the two complexes, indicating that p50 is present in both DNA-protein complexes (Fig. 1B, lane  10). The anti-p65 antibody had no effect on the faster DNAprotein complex, whereas it supershifted the slower migrating complex (Fig. 1B, lane 11). Antibodies to p52, c-Rel, and RelB had no effect (data not shown). Similar supershifts were obtained upon incubation of nuclear extracts from TNF-␣-treated cells with the p50 and p65 antibodies (Fig. 1B, lanes 13 and 14), indicating that ET-1 and TNF-␣ induce the formation of the same binding complexes. The effect of ET-1 was time-dependent ( Fig. 2A), being maximal after 15 min and persisting for 3 h. Activation of NF-B by TNF-␣ occurred more rapidly and was observed as soon as 2-5 min after addition of the cytokine (Fig. 2B).
Taken together, these results indicate that ET-1 and TNF-␣ increase the formation of two NF-B binding complexes in human myofibroblastic HSC, one containing the p50/p50 homodimer and the other the p50/p65 heterodimer.
ET-1 and TNF-␣ Induce the Degradation of IB-␣ in Human Myofibroblastic HSC-We studied IB regulation by ET-1 and TNF-␣ in the cytoplasm of human HSC treated with ET-1 or TNF-␣ for various times. The levels of IB-␣ were analyzed by Western blot analysis using a specific IB-␣ antibody. A protein with an expected size of 37 kDa was detected in cytoplasmic extracts of unstimulated cells (Fig. 3, A and B). Treatment of human HSC with ET-1 resulted in degradation of IB-␣ after 15 min, in keeping with the time course of NF-B activation; total resynthesis was observed upon 3 h of treatment (Fig. 3A). TNF-␣-induced degradation of IB-␣ was observed within 2 min and was total after 5 min (Fig. 3B), in agreement with the rapid activation of NF-B by TNF-␣ (Fig. 2B). Since resynthesis of IB-␣ occurred when activation of NF-B was maintained, we investigated whether degradation of IB-␤ could account for this prolonged activation of NF-B. However, no signal corresponding to the 47-kDa protein IB-␤ was found in untreated human HSC, whereas IB-␤ was detected in untreated Jurkat cells taken as control (Fig. 3C).
These results demonstrate that both ET-1 and TNF-␣ induce the degradation of IB-␣, an event that is associated with the increased formation of NF-B complexes. The activation of NF-B does not involve IB-␤.
ET-1 and TNF-␣ Up-regulate Cyclooxygenase-2 (COX-2) via Activation of NF-B in Human Myofibroblastic HSC-We have previously shown that the growth inhibitory effects of ET-1 are mediated by prostaglandins in human myofibroblastic HSC (6). Therefore, we next explored whether ET-1 and TNF-␣ regulate COX-2 in these cells, through the NF-B site in the promoter of this gene (13).
Endothelin-1 caused a rapid and sustained up-regulation of COX-2 mRNA, which was maximally increased by 4-fold after 45 min and remained elevated for at least 18 h (Fig. 4A). The induction of COX-2 mRNA was associated with an increase in COX-2 protein, observed after 3 h stimulation with ET-1 and lasting for at least 18 h (Fig. 4B). Activation of COX-2 protein expression was reproduced by the ETB agonist sarafotoxin S6C (see Fig. 5B) and was totally blocked by the non-selective ETA/ ETB antagonist PD-142893 (not shown). In contrast, the selective ETA antagonist B-Q123 did not affect COX-2 stimulation by ET-1 (not shown). Finally, pertussis toxin treatment did not modify ET-1-induced COX-2 expression (not shown). Tumor necrosis factor-␣ also up-regulated COX-2 mRNA and protein expression with a time course similar to that of ET-1 (Fig. 4, A  and B), whereas COX-1 expression was not affected (Fig. 5C). It should be noted that, in contrast to many other cell types (12,21,22), COX-2 protein is expressed in untreated HSC (Figs. 4   FIG. 1. Activation of NF-B by ET-1, sarafotoxin S6C, 9 and 12), of p50 antibodies (19) (lanes 10 and 13), or of p65 antibody (lanes 11 and 14). EMSA was performed as described in A. Antibody supershifts, produced by binding of the p50 and p65 antibody, are identified by double arrowheads. The autoradiograms shown are representative of three experiments. and 5).
We investigated whether NF-B is involved in up-regulation of COX-2 by ET-1 and TNF-␣ by means of two NF-B inhibitors, the proteasome inhibitor MG-132 (23,24) and the antiinflammatory compound sodium salicylate (25). Both NF-B inhibitors effectively reduced activation of NF-B and IB-␣ degradation by TNF-␣ (Fig. 5A) and ET-1 (not shown). Sodium salicylate and MG-132 blunted activation of COX-2 mRNA (Fig. 5A), COX-2 protein (Fig. 5B) by ET-1, S6C, and TNF-␣. The NF-B inhibitors also prevented activation of COX activity, measured as PGE 2 released in the presence of exogenously added arachidonic acid (Fig. 6A). The expression of COX-1 protein was unaffected by these treatments (Fig. 5C). Finally, ibuprofen, an anti-inflammatory compound that inhibits cyclooxygenase (26,27) without affecting NF-B (25), had no effect on the induction of COX-2 protein expression (not shown).
COX activity was also measured as PGE 2 released from endogenous arachidonic acid. As shown in Fig. 6B, in these conditions, ET-1 also caused a 3-fold increase in PGE 2 release, which was prevented by SC-58125, an inhibitor of COX-2 activity (28) (Fig. 6B). Similar results were obtained when assessing the effect of SC-58125 on stimulation of COX activity by ET-1 or TNF-␣ in the presence of exogenous arachidonic acid (not shown).
We also used dexamethasone, which inhibits COX-2 transcription by blocking activation of the transcription factors NF-B and NF-IL6 present in the COX-2 promoter and also by inducing a rapid destabilization of COX-2 mRNA (12). It should be noted that, as described in recent studies (29), dexamethasone treatment did not modify either IB-␣ degradation or NF-B DNA binding activation by ET-1 and TNF-␣ (not shown), suggesting direct interference of dexamethasone with the transactivation potential of NF-B (29). Dexamethasone reduced ET-1-stimulated COX-2 protein expression to levels below that of untreated cells (Fig. 6B, inset). Stimulation of PGE 2 release by ET-1 was also abolished following treatment with dexamethasone (Fig. 6B). Finally, ibuprofen, a COX-1/ COX-2 inhibitor, reduced stimulation of similar release by ET-1 (Fig. 6B) to levels comparable to those obtained with dexamethasone and SC-58125. It should be stressed that basal PGE 2 release (Fig. 6B) and basal COX activity (not shown) were dramatically reduced by dexamethasone and by the selective COX-2 inhibitor SC-58125, in keeping with the basal expression of COX-2 in human HSC (Figs. 4 and 5). Moreover, ibuprofen had a similar inhibitory effect on basal PGE 2 release than the COX-2 inhibitors (Fig. 6B). These results indicate that the major part of PGE 2 released in basal conditions derives from COX-2 activity.
Altogether these results indicate that not only COX-1 but also COX-2 is constitutively expressed in human myofibroblastic HSC and that the COX-2 isoform accounts for the constitutive COX activity. They also demonstrate that in human HSC, ET-1 and TNF-␣ up-regulate COX-2, and this induction involves NF-B activation.

The Inducible Cyclooxygenase-2 (COX-2) Mediates the Growth Inhibitory Effect of ET-1 and TNF-␣ in Human
Myofibroblastic HSC-We have shown that ET-1 inhibits the growth of human myofibroblastic HSC (5, 6), but the effect of TNF-␣ has not yet been investigated. As shown in Fig. 7B, TNF-␣ dose-dependently inhibited the growth of human myofibroblastic HSC. This contrasts with the mitogenic effect of the cytokine reported in rat HSC (30). We next investigated whether induction of COX-2 is involved in the regulation of human myofibroblastic HSC proliferation. Addition of SC-58125 or dexamethasone to serum-stimulated cells strongly reduced the antiproliferative effects of ET-1 (Fig. 7A) and TNF-␣ (Fig. 7B). These results suggest that activation of COX-2 is a key event in the pathway leading to inhibition of HSC proliferation by ET-1 and TNF-␣.
It should be noted that blocking COX-2 activity with either SC-58125 or dexamethasone enhanced [ 3 H]thymidine incorporation into DNA of serum-stimulated cells (see legend to Fig. 7). Accordingly, serum by itself enhanced PGE 2 production (not shown), which in turn inhibits myofibroblastic HSC growth (6). Therefore, although mitogenic, serum also generates negative FIG. 2. Kinetics of NF-B activation by ET-1 and TNF-␣ in human myofibroblastic HSC. Nuclear extracts were prepared from HSC made quiescent by incubation in serum-free medium over 3 days and further incubated for various times with 0.1 M ET-1 (A) or 50 ng/ml TNF-␣ (B). EMSA was performed as described under "Experimental Procedures" using a radiolabeled probe containing the B motif of the HIV-LTR promoter. The autoradiogram shown is representative of four experiments.

FIG. 3. Kinetics of IB-␣ degradation by ET-1 and TNF-␣ in human myofibroblastic HSC.
Cytoplasmic extracts were obtained from confluent HSC made quiescent by incubation in serum-free medium over 3 days and further incubated for various times with 0.1 M ET-1 (A) or 50 ng/ml TNF-␣ (B). Western blot measurements of IB-␣ in the cytoplasmic extracts were performed as described under "Experimental Procedures" using a specific IB-␣ antibody. C shows a Western blot of cytoplasmic extracts obtained from untreated HSC and Jurkat cells and was revealed with a specific IB-␤ antibody (19). signals for HSC proliferation. Similar findings were obtained when using other mitogens, such as platelet-derived growth factor-BB or thrombin. 2

DISCUSSION
We show here that ET-1 and TNF-␣, two factors that inhibit the growth of human myofibroblastic HSC, activate NF-B in these cells. Moreover, we demonstrate that activation of NF-B by ET-1 and TNF-␣ results in the induction of COX-2 and leads to growth inhibition of myofibroblastic HSC. These data suggest that NF-B may play an important role in the negative regulation of human myofibroblastic HSC proliferation.
Our results constitute the first demonstration that ET-1 increases the formation of NF-B complexes that we identified as p50/p50 and p50/p65 dimers; identical complexes were also enhanced by TNF-␣. Stimulation of NF-B DNA binding by ET-1 involves the growth inhibitory endothelin B receptor, since this effect is reproduced by the ETB receptor agonist sarafotoxin S6C. The increase in NF-B binding correlates with the rapid and transient degradation of IB-␣ in the cytosol of human HSC treated by ET-1 or TNF-␣. Since IB-␤ is not detected in human HSC, our results suggest that degradation of IB-␣ may initiate translocation of NF-B into the nucleus.
Other stimuli that enhance NF-B in HSC are poorly characterized, but NF-B-binding activity is increased during the activation process of rat HSC triggered by oxidative stress (31). Moreover, whereas classical activators of NF-B include proinflammatory cytokines, viruses, and physical stress (7-9), stimulation by agonists that bind to G-protein-coupled receptors

FIG. 4. Induction of COX-2 mRNA and protein by ET-1 and TNF-␣ in human myofibroblastic HSC.
A, induction of COX-2 mRNA by ET-1 and TNF-␣. HSC made quiescent by incubation in serum-free medium over 3 days and further incubated for various periods with 0.1 M ET-1 or 50 ng/ml TNF-␣, and total RNA was prepared. Northern blot analysis was performed using 20 g of total RNA and hybridized with a 32 P-labeled cDNA probe encoding COX-2 and G3PDH, as described under "Experimental Procedures." Typical autoradiograms of COX-2 signal are shown, which were quantified relative to G3PDH. The same results were obtained in two or three different experiments. B, Western blot analysis of COX-2 protein.
Whole cell extracts obtained from quiescent HSC treated with ET-1 or TNF-␣ for various periods. A typical autoradiogram is shown that was reproduced three times with similar results. EMSA was performed using nuclear extracts from quiescent HSC treated for 60 min with or without 50 ng/ml TNF-␣ in the absence or in the presence of MG-132 or sodium salicylate. Western blot measurement of IB-␣ was performed on the cytoplasmic extracts from the same cells, as described in Fig. 3. Right panel, COX-2 mRNA induction by ET-1 and TNF-␣. Total RNA was prepared from quiescent HSC treated for 3 h with ET-1 or TNF-␣ in the absence or in the presence of MG-132 or sodium salicylate. Northern blot analysis was performed as described in Fig. 4. B, COX-2 protein induction by ET-1 and TNF-␣. Whole cell lysates were prepared as follows: upper part, quiescent HSC pretreated with or without 50 M MG-132 for 60 min and further incubated with 0.1 M S6C, ET-1, or 50 ng/ml TNF-␣ for 8 h; lower part, quiescent HSC pretreated with or without 20 mM sodium salicylate for 30 min and further incubated with S6C, ET-1, or TNF-␣ for 3 h. Western blot analysis was performed as described in Fig.  4, using a selective COX-2 antibody (21). C, effect of ET-1 and TNF-␣ on COX-1 protein induction. Whole cell extracts and Western blot analysis were performed as in B, using a selective COX-1 antibody (20). In each case a typical autoradiogram is shown that is representative of results obtained in two or three experiments. has barely been reported (32)(33)(34)(35)(36). The finding that ET-1 and TNF-␣ activate NF-B in human myofibroblastic HSC constitutes another example of convergent pathways activated by growth inhibitory signals provided by cytokines and G-protein coupled-receptors. The signaling cascade activated upstream of IB-␣ remains to be determined but may involve one of the recently identified IB kinases (37)(38)(39)(40).
Among the possible target genes of NF-B in human myofibroblastic HSC, we focused on type 2 cyclooxygenase, since the promoter of this gene contains two NF-B consensus sites in its promoter region (13), and because we have demonstrated that inhibition of human HSC growth involves prostaglandin production (6). However, the presence and regulation of COX isoforms have not been previously documented in HSC. Our results demonstrate that the two COX isoforms, COX-1 and COX-2, are present in human HSC. As described in other cells, COX-1 is constitutively expressed in human HSC, but, in contrast to many other cell types, unstimulated HSC also express the COX-2 protein isoform. Activation of COX activity by ET-1 and TNF-␣ is reduced by dexamethasone and by the selective COX-2 inhibitor SC-58125 to levels comparable to that obtained with the COX-1/COX-2 inhibitor ibuprofen. These results, together with the fact that ibuprofen, SC-58125, and dexamethasone dramatically reduce basal COX activity, indicate that, in human HSC, the major part of PGE 2 levels found in basal conditions derives from COX-2 activity. Therefore, COX-2 accounts for the constitutive COX activity. The role of the "constitutive" COX-1 isoform in human HSC remains to be explored, in particular in view of recent results showing that COX-1 may behave as a delayed response gene (41). The COX-2 isoform present in human HSC is also inducible, as shown by the rapid increase of its mRNA and protein expression in response to ET-1, sarafotoxin S6C, and TNF-␣. The equal potency of ET-1 and sarafotoxin S6C to activate COX-2 (Fig. 5) and the lack of effect of the ETA antagonist on ET-1-induced COX-2 expression (not shown) indicates that the ETB receptor is involved. Up-regulation of COX-2 mRNA by these factors is sustained and is associated with a long-lasting increase in protein and activity. Induction of COX-2 by ET-1 and TNF-␣ is blunted by two different classes of NF-B blockers, the proteasome inhibitor MG-132, which blocks IB-␣ degradation (23,24), and the anti-inflammatory compound sodium salicylate (25), which is a potent inhibitor of NF-B, while being ineffective on purified cyclooxygenase activity (26). These NF-B inhibitors do not affect COX-1 protein expression. These results, which indicate that up-regulation of COX-2 requires NF-B activation, are consistent with the presence of an NF-B regulatory element in the promoter of the COX-2 gene (13), which is absent from that of the COX-1 gene (12). Our data are in line with recent observations implicating NF-B in COX-2 activation by interleu- kin-1, TNF-␣, or lipopolysaccharide by means of p65 antisense or NF-B inhibitors (13,(42)(43)(44).
Results from several studies indicate that COX-2 may be involved in the regulation of cell proliferation. Whereas most reports support a role of COX-2 in mitogenesis (45,46), recent data have shown that COX-2 may also be involved in growth arrest (47,48). Interestingly, Bornfeldt et al. (47) have recently reported that overexpression of COX-2 in human smooth muscle cells is associated with a marked inhibition of proliferation. In human HSC, blunting COX-2 activity with dexamethasone and with the selective COX-2 inhibitor SC-58125 blocks the growth inhibitory effect of ET-1 and TNF-␣, indicating that activation of COX-2 by either factor is a key event in the pathway leading to inhibition of HSC proliferation. We have previously shown that rapid production of prostaglandins leads to an increase in cyclic AMP and to the blockade of early events preceding human HSC proliferation such as mitogen-activated protein kinase activation (6). Our preliminary results suggest that this rapid production of prostaglandins would be ensured by stimulation of the COX-2 activity expressed in resting myofibroblastic HSC. 3 In contrast, induction of COX-2 mRNA and protein by ET-1 and TNF-␣ would result in a more sustained production of prostanoids and could be involved in the inhibition of more distal events, such as regulation of ET-1 receptors 2 and regulation of cell cycle components (49).
As described for COX-2, activation of NF-B has generally been linked to mitogenesis (50 -54), but recent data indicate that induction of NF-B is also associated with growth arrest and cellular differentiation (14 -16). In particular, NF-B stimulates the tumor suppressor gene p53 (16), and p65 interacts with the cell cycle inhibitor p21 WAF1 . (15). Also, overexpression of the NF-B member c-Rel arrests the proliferation of HeLa cells by interacting with p21 WAF1 (14). Because of the cytotoxic effect of NF-B inhibitors during prolonged cell growth experiments, the present study does not provide a direct demonstration for a role of NF-B in the negative regulation of human HSC growth. Nevertheless, it is tempting to speculate that NF-B, by up-regulating COX-2 in response to ET-1 and TNF-␣, plays a major role in the negative regulation of human myofibroblastic HSC growth.
The physiopathological consequences of our findings in the context of liver fibrosis remains to be determined. On the one hand, the consensus sequence for NF-B is present in the promoter of several genes activated by TNF-␣, which are involved in the inflammatory process associated with the development of liver fibrosis. For example, monocyte chemotactic protein-1 or intercellular adhesion molecule 1 are expressed in rat HSC and up-regulated by TNF-␣ (55, 56); however, TNF-␣ is mitogenic for rat HSC (30). On the other hand, prostaglandins E are antifibrogenic in an in vivo model of liver fibrosis (57), and in vitro studies show that in rat HSC, prostaglandins E inhibit both collagen I production (57) and cell proliferation (58). These findings, together with the fact that ET-1 and TNF-␣ inhibit the growth of human myofibroblastic HSC, following NF-B-dependent COX-2 induction, would be in favor of a negative regulatory role of this pathway in liver fibrogenesis.
In summary, we show here that two growth inhibitory peptides for human myofibroblastic HSC, ET-1 via its ETB receptors and TNF-␣, induce NF-B-dependent up-regulation of COX-2, resulting in inhibition of HSC proliferation. These results suggest that NF-B may play an important role in the negative regulation of human myofibroblastic HSC proliferation.