UV Irradiation Activates JNK and Increases αI(I) Collagen Gene Expression in Rat Hepatic Stellate Cells*

Hepatic stellate cells (HSCs) become activated into myofibroblast-like cells during the early stages of hepatic injury associated with fibrogenesis. The subsequent dysregulation of αI(I) collagen gene expression is a central pathogenetic step during the development of cirrhosis. Our recent study in rat HSCs (Davis, B. H., Chen, A., and Beno, D. (1996) J. Biol. Chem. 271, 11039–11042) found that ERK1,2 activation might be required for maximal αI(I) collagen gene expression. However, the role of the parallel JNK cascade in regulating αI(I) collagen gene expression was unknown. In this study, we initially found that UV irradiation of HSCs activated JNK but not ERK1,2. Furthermore, UV irradiation increased endogenous α I(I) collagen mRNA abundance and stimulated α I(I) collagen gene transcription in HSCs. The effect of the activation of JNK and Jun on α I(I) collagen gene expression was further evaluated via transfection of chloramphenicol acetyltransferase reporter plasmids with various sizes of truncated 5′ upstream promoter sequence (UPS) of the αI(I) collagen gene. This revealed that dominant negative transcription factor JUN suppressed α I(I) collagen gene transcription in HSCs maintained in media with 20% serum and constitutively activated JUN increased αI(I) collagen gene transcription in HSCs cultured in media with 0.4% serum. UV activated JNK utilized a distal GC box in the 5′-UPS of the collagen gene to regulate gene transcription. This observation was confirmed by site-directed mutagenesis. In co-transfection experiments, the col-chloramphenicol acetyltransferase reporter with a mutagenized GC box was not suppressed by dn-JUN and was not stimulated by activated JUN or by UV irradiation. Southwestern blotting analyses and gel shift assays with basic transcription element-binding protein antiserum suggested that the GC box was bound by basic transcription element-binding protein, a recently described DNA-binding protein. In conclusion, the current study combined with our previous report suggests that ERK1,2 and JNK cascades regulate αI(I) collagen expression in HSCs through different regions of the 5′-UPS of the gene. The distal GC box in the 5′-UPS of the αI(I) collagen gene may play a central role in receiving extracellular signals through the JNK pathway.

Eukaryotic cells have developed specific signal transduction pathways to respond to and integrate extracellular stimuli. Three of these pathways that have been elucidated in eukaryotic cells can be simplified as follows: 1) raf 3 MEK1,2 3 ERK1,2; 2) MEKK1 3 SEK1 3 JNK/SAPK; and 3) MEKK1 3 SEK1 3 p38 (1). In each of these kinase cascade pathways, the upstream kinase phosphorylates and activates its immediate downstream substrate kinase. Extracellular signals are thereby transduced through these cytoplasmic kinase cascades to reach their nuclear targets and regulate gene expression. Recent studies indicate that the raf 3 ERK1,2 pathway has significant effects on ␣I(I) collagen gene expression in ratderived hepatic sinusoidal stellate cells (HSCs), 1 the major effector cells during the overproduction of collagen which typifies hepatic fibrogenesis and cirrhosis (2)(3)(4)(5). The response elements for the cascade involved a NF-1 site in the proximal promoter of the collagen gene, as well as a region within Ϫ1620 to Ϫ1630 in the distal promoter of the ␣I(I) collagen gene (2). The importance of the NF-1 site in regulating collagen gene expression is in agreement with other studies using non-HSCs (6 -9). The effects of JNK activation on ␣I(I) collagen gene expression had not been evaluated in HSCs. Activated JNK can phosphorylate and activate its target protein, JUN, and transduce a signal from cytoplasm to nucleus (10 -13). JNKs are activated by a variety of stimuli, such as UV irradiation, heat shock, and osmotic imbalance (14,15). HSCs have been shown to contain the JNK cascade, which can be activated by fibronectin, interleukins, and tumor necrosis factor (16). These are classic compounds associated with tissue injury and may also be present in serum (16). The current report studied the effects of UV irradiation of HSCs on activation of JNK and on ␣I(I) collagen gene expression. It was found that exposure of HSCs to UV light activated JNK but not ERK1,2, and the activation of JNK by UV irradiation increased ␣I(I) collagen mRNA abundance. Further studies indicated that activated JNK regulated ␣I(I) collagen gene expression through a distal GC box located in the 5Ј-UPS of the ␣I(I) collagen gene. This response element was distinct from that utilized by the ERK1,2 cascade. A 32-kDa protein, designated basic transcription element-binding protein (BTEB), was found to bind to the GC box. BTEB DNA binding activity was up-regulated in activated stellate cells. The potential mechanism(s) utilized by UV-induced activation of JNK to stimulate ␣I(I) collagen gene transcription is further discussed.

MATERIALS AND METHODS
Cell Culture and Transient Transfection-Hepatic stellate cells (HSCs) were isolated from Sprague-Dawley male rats and subcultured in DMEM supplemented with 10% fetal bovine serum and 10% newborn calf serum (10/10 DMEM) by previously described methods (17,18). Experimental manipulations were performed with cells at passages 2-6. Sixty to eighty percent confluent cells in 6-well cell culture plates were transfected by the LipofectAMINE method following the protocol provided by the manufacturer (Life Technologies, Inc.). Equimolar col-CAT reporters (1-1.3 g of DNA) and equal amounts of dominant negative JUN or empty control plasmid pMNC (2 g) were used in each well. For each transfection, 0.2 g of ␤-galactosidase expression plasmid pSV-␤-gal (Promega) was co-transfected for evaluation of transfection efficacy. Each DNA transfection experiment was performed in triplicate. Each transfection was repeated at least three times.
Chemicals and UV Irradiation-Unless specifically noted, all chemicals were purchased from Fisher or Sigma. Seventy to eighty percent confluent passaged HSCs were incubated in DMEM with 0.4% fetal bovine serum for 48 h before exposure to a UV lamp (254 nm, 38 W, 76-cm distance between uncovered plates and the UV lamp) in a tissue culture hood for 0.2 or 0.4 min. The UV dose was approximately 10 J/m 2 (19). Control cells were exposed to regular light in a tissue culture hood for 0.4 min. Cells were then maintained in the media with 0.4% serum in a 37°C incubator for an additional 1 h before total RNA isolations or protein extractions.
RNA Isolation and RNase Protection Assay (RPA)-Total RNA was isolated by the TRI-Reagent (Sigma) following the protocol provided by the manufacturer. Single-stranded RNA probe complimentary to ␣I(I) collagen mRNA nucleotides was made from the first exon (1-206) of the gene via polymerase chain reaction (PCR). Primers for the PCR were 5Ј-CGGGATCCCGAGCAGACGGGAGTTTCACC-3Ј (the boldface region generated a BamHI site) and 5Ј-TCCCCCGGGGGAGAACT-TACTGTCTTCTTGG-3Ј (the boldface region generated a SmaI site). The PCR product was then subcloned into BamHI and SmaI sites in the plasmid pGEM-3Zf(ϩ) (Promega). The T 7 promoter in the plasmid was used to generate single strand antisense RNA probe. The template for rat cyclophilin was obtained from Ambion and yields a 103-base pair protected fragment. The antisense RNA probes were synthesized and labeled by MAXIscript in vitro transcription kits (Ambion). The synthesized probes were gel purified. RPA was carried out with RPA II (Ambion) following the protocol provided by the manufacturer. The dried gel was exposed to a phosphor imaging system (Phosphor Image SI, Molecular Dynamics, Sunnyvale, CA). The radioactivity in each band was measured by computer-aided densitometry of the phosphorimage using IPLab Gel (Signal Analytics Corp.) as described previously (20).
CAT Assay and Transfection Efficacy Normalization-Cells were harvested and assayed for CAT activity as described previously (2). In brief, cells were washed twice and harvested in cold PBS by cell scrapers. Cells in PBS were lysed by three cycles of freeze-thaw. After centrifugation for 15 min at high speed at 4°C, protein concentrations were determined by QuantiGold (Diversified Biotech). Protein extracts were heated for 5 min at 65°C to inactivate any endogenous acetylases. The protein extracts reacted with 50 l of CAT assay mixture at 37°C for 90 min. The mixture contains 350 l of PBS, 110 l of 10 mM acetyl-CoA, 1.5 ml of chloramphenicol, and 100 l of 0.5 mCi/ml [ 3 H]acetyl-CoA. The reaction mixtures were added to Econofluor-2 (Packard Inc.) in scintillation vials. The CAT activity was analyzed in a liquid scintillation analyzer. Transfection efficacy was normalized by analyzing co-transfected ␤-galactosidase activity expressed as ␤-gal relative unit/mg protein by utilizing Galacto-Light chemiluminescent reporter ␤-galactosidase assay kit (Tropix). Finally, the CAT activity of each transfection was expressed as relative units/mg of protein after normalization of transfection efficacy as calculated by ␤-galactosidase activity. Transfection efficiency was estimated by counting blue cells as described by Promega. Cells transfected with the ␤-gal reporter plasmid were stained blue by x-gal substrate after cell fixation, whereas untransfected cells remained unstained. Transfection efficiency was estimated by selecting 10 random fields on the microscope and counting the percentage of blue cells versus unstained cells.
Plasmid Constructions-Dominant negative JUN (dn-JUN) was originally from Dr. M. J. Birrer (10) The constitutively active form of JUN (v-JUN) was a gift from Dr. N. Hay (University of Chicago). Both dn-JUN and v-JUN were tested; the expected results were obtained when an AP-1 reporter plasmid 3x-TRE-CAT was used, and there was no effect on pBL-CAT. Plasmid 3x-TRE-CAT contains three AP-1 sites to regulate CAT gene expression. The control empty parental vector pBL-CAT has no AP-1 sites. Both plasmids were kindly provided by Drs. B. J. Aneskievich and E. Fuchs (21). The colCAT reporter plasmid p1.7/1.6 contains 1.7 kilobases of the 5Ј-UPS of the rat ␣I(I) collagen gene and 1.6 kilobases of the first exon and part of the first intron linked to the CAT reporter gene (22,23). Plasmids p1.3/1.6 and p0.4/1.6 col-CAT reporters were derivatives of plasmid p3.6/1.6 produced by digestion with NheI/TthIII I and NheI/MfeI restriction endonucleases, respectively. Plasmids p3.6/1.6 and pdel1.3-0.4/1.6 were as described previously (2). pdel 1.4 -0.4 was generated by PCR. The upstream primer (Ϫ1506 to Ϫ1487) was 5Ј-CACCTAGCTAGCGGAATCTTG-GATGGTTTGG-3Ј. Twelve additional nucleotides with an NheI restriction site were added to the 5Ј-end of the primer. The downstream primer (Ϫ1412 to Ϫ1429) was 5Ј-CCTCAATTCAGGCCATAGACGTC-CTGTATC. Twelve additional nucleotides with an MfeI restriction site were added 5Ј of the primer. The generated plasmid pdel 1.4 -0.4/1.6 was sequenced to ensure that no changes occurred during the PCR.
PCR, DNA Sequencing, and Site-directed Mutagenesis-PCR was performed using Ultma DNA Polymerase (Perkin-Elmer). The sequencing procedures followed the protocol of Sequenase Version 2.0 DNA Sequencing Kit from Amersham Pharmacia Biotech. Plasmid p1.7(GC box mut)/1.6 was created by the overlap extension method of sitedirected mutagenesis (24,25). The site-directed mutants were sequenced to confirm the mutations. The site (Ϫ1494 to Ϫ1468) was mutated from 5Ј-GGTTTGGAGGAGGCGGGACTCCTTGC-3Ј to 5Ј-GGTTTGGAGGAAATAAGACTCCTTGC-3Ј. This site contains the GC box of interest (i.e. Ϫ1484 to Ϫ1475) Nuclear Extraction-Nuclear proteins were prepared as follows. Ͼ95% confluent cells in cell culture flasks were washed twice and harvested in cold PBS. Cells were collected by centrifugation at 2000 rpm for 5 min at 4°C. The cell pellet was resuspended in 5-10ϫ volume of Solution A, which contained 10 mM Hepes, pH 7.9, 1.5 mM MgCl 2 , and 10 mM KCl, and incubated for 10 min at 4°C. Cells were re-collected by centrifugation for 5 min at 2000 rpm. The cell pellet was resuspended in 1.5-2 ml of Solution A with the following inhibitors: 0.5 mM dithiothreitol, 10 g/l leupeptin, 2 mM phenylmethylsulfonyl fluoride, 10 mM NaF, 10 g/ml aprotonin, 1 mM NaVO 3 , and 60 mM ␤-glycerophosphate. Cells were gently stroked 30 times on ice in a Dounce Type B homogenizer. Nuclei were collected by centrifugation at 2500 rpm for 10 min at 4°C. The nuclei pellet was resuspended in Solution A with the above inhibitors and subsequently centrifuged at 15,000 rpm for 20 min at 4°C. The pellet was resuspended in Solution C, containing 20 mM Hepes, pH 7.9, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol, 10 g/l leupeptin, 2 mM phenylmethylsulfonyl fluoride, 10 mM NaF, 10 g/ml aprotonin, 1 mM NaVO 3 , and 60 mM ␤-glycerophosphate. The nuclei suspension was incubated on an up-and-down rocker for 60 -90 min at 4°C and then centrifuged at 15,000 rpm for 30 min. The clear supernatant was aliquoted in microcentrifuge tubes and stored at Ϫ70°C until use. Nuclear protein concentrations were quantified by Bio-Rad reagent.
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear protein extracts (5-10 g) were incubated in 10 l of binding buffer with 4% glycerol, 1 mM MgCl 2 , 0.5 mM dithiothreitol, 50 mM NaCl, 10 mM Tris-HCl, pH 7.5, and 2.5 g of poly(dI-dC). In competition EMSAs, the indicated amounts of double-stranded oligonucleotides were added as unlabeled competitor. The mixture was incubated at room temperature for 10 min before adding a radiolabeled probe. In some cases, 2 l of anti-BTEB or preimmune serum was added to the mixture and incubated at room temperature for 20 min prior to adding 1 ng of a radiolabeled probe. Nucleoprotein complexes were resolved by electrophoresis on 6% nondenaturing polyacrylamide gels in 0.5x Tris borate/EDTA (TBE) buffer. The gels were exposed to autoradiography film (DuPont) overnight at Ϫ70°C. Double-strand oligonucleotide probes were radiolabeled by T 4 polynucleotide kinase (New England Biolabs) and [␥-32 P[dATP (DuPont). Polyclonal antibody ␣-BTEB was commercially generated by Research Genetics, Inc., according to the rat BTEB cDNA and its deduced amino acid peptide sequence (26).
Oligonucleotide Probes-The double-strand oligonucleotides used in the EMSA were as follows. SP-1 consensus oligonucleotides purchased from Promega: 5Ј-ATTCGATCGGGGCGGGGCGAGC-3Ј; GC box oligonucleotides from Ϫ1491 to Ϫ1470 of the 5Ј-UPS of the rat ␣I(I) collagen gene: 5Ј-TTGGAGGAGGCGGGACTCCTTG-3Ј. Underlined sequences represent GC box binding domain. All of the oligonucleotide probes were synthesized by Life Technologies, Inc.
Southwestern Blot Analysis-Nuclear extracts (100 g/lane) were obtained either from HSCs cultured in DMEM with 20% or 0.4% serum for 48 h or directly from normal uncultured rat HSCs or rats pretreated with carbon tetrachloride (CCl 4 ) (0.5 ml ϩ 0.5 ml of mineral oil given intraperitoneally) 16, 24, 48 or 72 h prior to sacrifice. The extracted proteins and prestained protein standards were resolved by 15% SDSpolyacrylamide gel electrophoresis and transblotted to a nitrocellulose membrane filter. The filter was washed for 30 min in Buffer A (25 mM NaCl; 25 mM Hepes-NaOH, pH 7.9; 5 mM MgCl 2 ; 0.5 mM dithiothreitol, which was added fresh prior to use). The filter was then blocked by incubation for 3 h at room temperature in Buffer A with 5% nonfat milk. After a short wash with Buffer A containing 0.25% nonfat milk, the filter was probed by 32 P-labeled double-strand GC box oligonucleotide (from Ϫ1491 to Ϫ1470 of the 5Ј-UPS of the rat ␣I(I) collagen gene) 5Ј-TTGGAGGAGGCGGGACTCCTTG-3Ј in 10 ml of Buffer A with 5 g/ml poly(dI-dC) plus 0.25% nonfat milk on a rocker at 4°C overnight. The filter was briefly washed twice with Buffer A containing 0.25 M NaCl and then washed with regular Buffer A twice at room temperature for 15 min each. The filter was wrapped and exposed to autoradiography film (DuPont) overnight at Ϫ70°C.
Statistics-Differences between means were analyzed via Student's t test using the Statworks program for the Macintosh. Statistical significance required differences at the level of p Ͻ .05.

JNK Activation by UV Irradiation Increases Endogenous
␣I(I) Collagen Gene Expression-The effects of activation of JNK on ␣I(I) collagen gene expression were studied in passaged HSCs treated with UV irradiation (Fig. 1). HSCs were grown to near confluence and then incubated for 48 h in serumdepleted medium (0.4% serum). The cells were then exposed to ultraviolet light for 0.2 or 0.4 min or to regular light for 0.4 min. The cell extracts were prepared for Western blot analysis (Fig.  1A). It was found that the active forms of JNK 1 and JNK 2 were readily detected in HSCs exposed to UV light by antiactivated JNK polyclonal antibody (Promega), which preferentially recognized the dually phosphorylated active forms of JNK enzymes (Fig. 1A). In contrast, faint bands of active forms of JNK were present in HSCs exposed to regular light (Fig. 1A). UV irradiation did not induce activation of MAPK (ERK1,2) in serum-starved HSCs probed with the anti-activated MAPK polyclonal antibody (Promega) (data not shown). To determine the effects of UV-induced activation of JNK on endogenous ␣I(I) collagen gene expression in HSCs, ␣I(I) collagen mRNA from serum-starved HSCs treated with or without UV irradiation was measured by RPA. (Fig. 1B). The radioactivity in each band in the RPA was measured and quantitated by computer-aided densitometry of the phosphorimage using IPLab Gel. As shown in Fig. 1C, it demonstrated that JNK activation by UV irradiation caused an approximate 2.9-fold increase in endogenous ␣I(I) collagen mRNA steady state levels in HSCs.
Activation of JUN Stimulates colCAT Expression in Cultured HSCs-To understand the mechanisms of UV induction of ␣I(I) collagen gene expression, a series of transfection experiments were performed in passaged HSCs (Fig. 2). HSCs were transfected with an AP-1 reporter plasmid 3x-TRE-CAT or an empty parental plasmid pBL-CAT ( Fig. 2A). The Ap-1 reporter plasmid 3x-TRE-CAT has three AP-1 binding sites to regulate CAT gene expression, whereas the pBL-CAT is the control parental plasmid without AP-1 binding sites. These two plasmids have been used previously to study activation of AP-1 and transcription mediated by AP-1 (27). After transfection, the cells were incubated in media containing 0.4% serum for 36 h before exposure to UV light or regular light. The results indicated that UV irradiation induced a Ͼ4-fold increase in the CAT activity in cells transfected with 3x-TRE-CAT ( Fig. 2A). In contrast, UV exposure did not change the CAT activity in HSCs transfected with pBL-CAT (Fig. 2A). These results suggested that UV irradiation increased the ability of the AP-1 complex to regulate gene transcription. To study the significance of the activation of JUN in stimulating ␣I(I) collagen gene expression, HSCs were co-transfected with colCAT plasmids and a v-JUN expression plasmid (Fig. 2B) or a dn-JUN plasmid (Fig. 2C). In Fig. 2B, the results illustrated that v-JUN, but not empty control plasmid pMNC, increased the CAT activity in serumstarved HSCs co-transfected with the p1.7/1.6 colCAT reporter plasmid, which contained 1700 base pairs of the 5Ј-UPS and 1600 base pairs of the first exon and part of the first intron of the ␣ I(I) collagen gene linked to a CAT reporter gene. However, v-JUN did not increase the CAT activity in HSCs cotransfected with the p0.4/1.6 colCAT plasmid (Fig. 2B). Fig. 2C showed that dn-JUN had an opposite effect on p1.7/1.6 in HSCs maintained in medium with 20% serum. dn-JUN significantly

FIG. 1. UV irradiation activated JNK and increased endogenous ␣I(I) collagen gene expression in HSCs.
A, serum-starved HSCs were exposed to UV light for 0.2 min or 0.4 min, or to regular light for 0.4 min as control (ctr) before incubation in medium with 0.4% serum for an additional 1 h. Fifteen micrograms of total cell lysate proteins per lane were separated in 10% SDS-polyacrylamide gel electrophoresis and analyzed by Western blots with ANTI-ACTIVE JNK polyclonal antibody. B, 10 g/lane of total RNA from cells exposed to UV light (UV) or to regular light (no UV) for 0.4 min were analyzed by RPA. Lane P is the undigested control RNA probes. Arrows at the right indicate protected ␣I(I) collagen and cyclophilin mRNA. Representative assays are shown here. C, the radioactivity in each band in B was quantitated by computer-aided densitometry of the phosphor image using IPLab Gel. Cyclophilin was used as an internal control to normalize the loading of RNA in each lane. reduced the CAT activity in HSCs co-transfected with p1.7/1.6 but not in HSCs co-transfected with p0.4/1.6. The lack of the inhibitory effect in HSCs co-transfected with p0.4/1.6 suggests that the necessary response element is missing and that the dn-JUN inhibitory effect is not due to a nonspecific sequestration effect. It should also be emphasized that v-JUN experiments were carried out in 0.4% serum conditions to eliminate serum derived sources of JUN stimulation, whereas the dn-JUN experiments were done in the presence of serum (20%) as a source of JUN stimulation. Taken together, these data indicate that activation of JUN plays an important role in increasing ␣ I(I) collagen gene transcription in vitro and that the response element for activated JUN is located between Ϫ1.7 and Ϫ0.4 kilobases of the 5Ј-UPS of the ␣I(I) collagen gene.
The Distal GC Box in the 5Ј-UPS Is Required for JUN Stimulation-To further analyze the role of JUN in regulating ␣ I(I) collagen gene expression in HSCs, and to identify the promoter region responding to activated JUN, a series of colCAT reporter constructs were produced, some of which are shown in Fig. 3A, left. In order to identify the DNA response element, HSCs were co-transfected with dn-JUN and a series of colCAT plasmids with different 5Ј-UPS truncations of the ␣I(I) collagen gene, and the experiments were performed in DMEM with 20% serum (Fig. 3A, right). dn-JUN resulted in a 2.6-fold reduction in the CAT activity in cells co-transfected with p1.7/1.6 ( Fig. 3A and statistical analysis in Table I). A significant loss of responsiveness to the dn-JUN inhibitory effect was observed in cells co-transfected with p1.3/1.6 or p0.4/1.6 ( Fig. 3A and statistical analysis in Table I). Additional experiments were therefore focused on the general region between Ϫ1700 and Ϫ1300 base pairs. To prove that this region contained the required response element(s), additional constructs were produced, i.e. pdel 1.3-0.4/1.6 and pdel 1.4 -0.4/1.6 (see Fig. 3A for diagram). When transfected with the p0.4/1.6 reporter, there was no response to the co-transfected dn-JUN. In the pdel reporters, the region containing putative responsive element(s) is now adjacent to the unresponsive 0.4/1.6 region. When transfected with these pdel plasmids, the response to the dn-JUN was regained (see Fig. 3A). A computer-aided search revealed a GC box binding site in this region (Ϫ1475 to Ϫ1484). Because the GC box might be the element responding to dn-JUN, several key nucleotides in the GC box were mutated to evaluate their necessity for the dominant negative JUN inhibitory effect. As shown in Fig. 3A, this mutagenized reporter p1.7 (GC box mut)/1.6 lost the inhibitory response to dn-JUN. This result suggests that the distal GC box plays a key role in the response to the dn-JUN inhibitory effect on ␣I(I) collagen gene transcription. In order to confirm that the GC box was the response element mediating the regulation of ␣I(I) collagen gene expression by activated JNK, HSCs were co-transfected with the v-JUN and colCAT p1.7/1.6 or p1.7(GC box mut)/1.6 plasmid (Fig. 3B). v-JUN had a stimulatory effect on the CAT activity in cells co-transfected with wild-type colCAT p1.7/1.6. The cells co-transfected with p1.7(GC box mut)/1.6 lost the stimulatory effects of v-JUN on CAT expression. Further evidence of GC box function was provided by HSCs transfected with colCAT p1.7/1.6 or p1.7(GC box mut)/1.6 before exposure to UV irradiation (Fig. 3C). The col CAT plasmid with a mutated GC box (p1.7 (GC mut)/1.6) lost its response to stimulation by activated JNK. These results collectively suggest that the distal GC box in the 5Ј-UPS of the ␣I(I) collagen gene is the response element required for UV activated JNK to stimulate ␣I(I) collagen gene expression.
GC Box-binding Protein Is Altered During Stellate Cell Activation-The GC box sequence is one of the most common regulatory DNA elements of eukaryotic genes. A GC box is usually bound by a member of the Sp transcription factor family (e.g. Sp-1 transcription factor) or BTEB (26). To analyze the GC box-binding protein in HSCs, nuclear extracts were obtained from culture-activated HSCs and from HSCs isolated from rats injected with CCl 4 (72 h prior to sacrifice). Injection of CCl 4 results in HSC activation typified by HSC proliferation and ultimately enhanced ␣I(I) collagen gene expression (3)(4)(5). The extracts from cells activated in vitro and in vivo both contained a single GC box-binding protein, which was demonstrated in the gel shift assay as a single intense band with the same

FIG. 2. Activation of JUN-stimulated colCAT expression in HSCs.
A, HSCs were transfected with AP-1 reporter plasmid 3x-TRE-CAT or control plasmid pBL-CAT. The cells were incubated for 36 h in DMEM with 0.4% serum before exposure to UV light or to regular light for 0.4 min. B, HSCs were co-transfected with colCAT p1.7/1.6 or p0.4/1.6 and v-JUN or control plasmid pMNC. The cells were incubated in DMEM with 0.4% serum for 36 h before harvest for CAT assay. C, HSCs were co-transfected with colCAT p1.7/1.6 or p0.4/1.6 and dn-JUN or control plasmid pMNC. The cells were cultured for an additional 36 h in DMEM with 20% serum. The CAT activity was measured by CAT assay (see under "Materials and Methods" for details). Co-transfected ␤-galactosidase was used to normalize the transfection efficiency (n ϭ 6). mobility (Fig. 4, lanes 6 and 11, lower arrow). In contrast, a faint band with the same mobility was present in quiescent HSC extracts obtained directly from normal rats (lane 10) and in serum-deprived HSCs in culture (lane 5). Competition with either GC box oligonucleotides or a DNA fragment from Ϫ1400 to Ϫ1500 base pairs of the 5Ј-UPS, which contained the GC box eliminated this binding (lanes 7 and 8). Although Sp-1 binding could be involved in the GC box binding, this did not appear to be likely. As shown in Fig. 4, an Sp-1 consensus oligonucleotide resulted in a gel shift that differed considerably from the gel shift using the GC box probe (see Fig. 4, lanes 1-3 versus lanes  6 -11). Its mobility was slower, and the response to serumcontaining media was less significant. Also, competition with excess Sp-1 oligonucleotide had a minimal effect on the GC box gel shift (Fig. 4, lane 9). The GC box gel shift appeared to require 72 h following CCl 4 injection for maximal activity (see Fig. 5). Extracts from earlier time points were intact as they had the ability to bind to a consensus Sp-1 domain (data not shown). This gradual onset of the GC box-binding protein was consistent with the gradual increase in collagen gene expression in vivo (4). The process might require the recruitment of other cells and cytokine release to adequately stimulate the HSCs (28).
BTEB Is the GC Box-binding Protein-The results of the EMSA (Fig. 4) suggested that the DNA-binding protein is unlikely to be an Sp-1 transcription factor because the nucleopro- tein complex of interest moved much faster than the Sp-1 complex in the gel. Further studies were performed to identify the GC box DNA-binding protein in the 5Ј-UPS of the ␣I(I) collagen gene in HSCs. The Southwestern blot revealed a single 32-kDa protein, which could form a nucleoprotein complex with a GC box oligonucleotide (Fig. 6). The 32-kDa size is incorrect for Sp-1 (29). The band also became prominent at 72 h following CCl 4 injection (Fig. 6, lane 6), which was consistent with our previous observation (Fig. 5). As expected from the gel shift experiments, serum stimulation of cultured HSCs also resulted in a GC box-binding protein that had the same molecular mass (Fig. 6, lane 2). Recent studies involving other cell types unrelated to HSCs had identified and cloned a GC box-binding protein with a similar molecular mass referred to as BTEB (26,30). In another EMSA (Fig. 7), nuclear protein extract was pretreated with polyclonal anti-BTEB. The pretreatment with anti-BTEB significantly diminished the gel shift retarded band of interest (lane 3). In great contrast, pretreatment with preimmune serum had no effect on the abundance of the retarded band (lane 4). These gel shift assays suggest that BTEB is the distal GC box DNA-binding protein. DISCUSSION The sinusoidal HSCs represent the major effector cells during hepatic fibrogenesis. During this process, the normal quiescent, vitamin A-storing HSCs transform into actively proliferating, collagen-producing cells. The HSCs excessive collagen matrix production and increased cellular proliferation lead to the collagenization and disruption of the space of Disse and formation of the fibrous septae seen in cirrhosis. Eukaryotic cells have developed specific signal transduction pathways to   respond to and integrate extracellular stimuli. A recent study of HSCs found that fibronectin and tumor necrosis factor ␣ activated both JNK and ERK1,2 and increased AP-1 DNA binding ability and ␣ I(I) collagen mRNA abundance (16). However, that study could not clearly indicate the effects of JNK or ERK1,2 on ␣(I) collagen gene expression because fibronectin and tumor necrosis factor ␣ activated both pathways. Our previous studies found that activation of ERK1,2 induced by serum was important for the maximal expression of ␣I(I) collagen gene in passaged HSCs (2). The function of activated JNK in regulating ␣I(I) collagen gene expression was studied in the present report. It was found that UV irradiation activated JNKs, but not ERK1,2 (Fig. 1A) and that UV irradiation increased endogenous ␣I(I) collagen mRNA abundance in passaged HSCs (Fig. 1B). Both activated ERK1,2 and activated JNK have the capacity to translocate to the nucleus and phosphorylate transcription factors (31,32). The current transfection studies indicated that activated JNK induced by UV irradiation increased the ability of AP-1 to induce gene transcription in HSCs ( Fig. 2A). Additional studies suggested that the active form of JUN increased ␣I(I) collagen gene transcription (Fig. 2B) and, as expected, dominant negative JUN inhibited ␣I(I) collagen gene transcription in cultured HSCs (Fig. 2C). The response element mediating UV induction of ␣I(I) collagen gene expression was located in a distal GC box in the 5Ј-UPS of the collagen gene. A previous study found that an AP-1 binding site in the first intron of the ␣I(I) collagen gene played a critical role in the stimulation of the ␣I(I) collagen gene by transforming growth factor (TGF␤) (33). Another recent report indicated that the first intron of the ␣I(I) collagen gene had a tissue-specific and developmentally regulated function in transcription of the gene (34). In the present study, HSCs transfected with plasmid p1.7/1.6 or p1.7 (GC box mut)/1.6 showed significant differences in the CAT activity in their responses to dominant negative JUN (Fig. 3A), to the constitutively active form of JUN (Fig.  3B), and to UV irradiation (Fig. 3C). All of these experiments utilized plasmids containing the intact wild-type AP-1 binding site in the first intron. The only difference between the two plasmids was the distal GC box in the 5Ј-UPS. Five nucleotides inside the GC box were mutated by site-directed mutagenesis in the plasmid p1.7 (GC box mut)/1.6. The current study does not support the contention that the AP-1 binding site in the first intron is directly involved in the induction of ␣I(I) collagen gene transcription by exposure of HSCs to UV light. Further studies demonstrated that the GC box in the 5Ј-UPS of the ␣I(I) collagen gene was bound by BTEB, a recently described GC box DNA-binding protein. A GC box sequence could be bound by either Sp-1 or BTEB (26). BTEB does not share sequence similarity to Sp-1, except for a zinc finger domain of Cys-Cys/His-His motif that is repeated three times with 72% sequence similarity to Sp-1. The study by Imataka et al. (24) suggested that BTEB exerted different effects on gene transcription with differing numbers and positions of the GC box sequence in the promoter region. Further studies are needed to understand the relationship between BTEB functions and flanking nucleotides of the GC box. It is likely that BTEB binding to the GC box plays a role in regulating ␣I(I) collagen gene expression because 1) the GC box plays a contributory role in stimulating ␣I(I) collagen gene expression (see Fig. 3), 2) BTEB appears to be the major protein binding to the GC box at periods of increased ␣I(I) collagen gene expression (Fig. 4), and 3) the appearance of BTEB was coincident with activation of HSCs in both in vitro and in vivo settings as demonstrated by the results shown in Figs. 5 and 6. Additional studies are needed, such as co-transfections of HSCs with colCAT plasmids and BTEB overexpression construct and experiments involving BTEB overexpression and the measurement of endogenous collagen gene expression in stable transfectants in non-HSC models. There is little known about BTEB gene expression and regulation, and its relationship to fibrosis and liver injury remains unclear. One speculation is that activation of JNK and subsequent activation of JUN directly transduces signals to the promoter of the BTEB gene through JUN-JUN dimerization and then BTEB gene expression is stimulated. BTEB could then bind to the GC box in the 5Ј-UPS of the ␣ I(I) collagen gene and stimulate gene transcription. These speculations will require further studies.