Raf and mitogen-activated protein kinase regulate stellate cell collagen gene expression.

Hepatic stellate cells become activated into myofibroblast-like cells during the early stages of hepatic injury associated with fibrogenesis. The subsequent dysregulation of hepatic stellate cell collagen gene expression is a central pathogenetic step during the development of cirrhosis. The cytoplasmic Raf and mitogen-activated protein (MAPK) kinases were found to differentially regulate alpha I(I) collagen gene expression in activated stellate cells. This suggests an unappreciated branch point exists between Raf and MAPK. A MAPK-stimulatory signal was mapped to the most proximal NF-1 and Sp-1 binding domains of the 5'-untranslated region of the collagen gene. A Raf-inhibitory signal was mapped to a further upstream binding domain involving a novel 60-kDa DNA-binding protein (p60). The cell-specific expression and induction of p60 in stellate cells during the early stages of hepatic fibrogenesis in vivo suggest a central role for this pathway during liver injury and stellate cell activation.


Hepatic stellate cells become activated into myofibroblast-like cells during the early stages of hepatic injury associated with fibrogenesis. The subsequent dysregulation of hepatic stellate cell collagen gene expression is a central pathogenetic step during the development of cirrhosis. The cytoplasmic Raf and mitogen-activated protein (MAPK) kinases were found to differentially regulate ␣I(I) collagen gene expression in activated stellate cells. This suggests an unappreciated branch point exists between Raf and MAPK. A MAPK-stimulatory signal was mapped to the most proximal NF-1 and Sp-1 binding domains of the 5-untranslated region of the collagen gene. A Raf-inhibitory signal was mapped to a further upstream binding domain involving a novel 60-kDa DNAbinding protein (p60). The cell-specific expression and induction of p60 in stellate cells during the early stages of hepatic fibrogenesis in vivo suggest a central role for this pathway during liver injury and stellate cell activation.
The detailed cytoplasmic signaling involved in the regulation of key structural and/or disease-related genes is poorly understood. This information would be useful to (i) identify potential targets for therapeutic intervention and (ii) identify unappreciated DNA-binding domains involved in gene regulation, as well as to link these domains to potential upstream regulators. Previous studies have suggested that Ras overexpression leads to a decrease in type I collagen gene expression (1). This implies that cytoplasmic kinase cascades may be involved in the regulation of the collagen gene. This gene has physiologic significance as it plays a major role in embryogenesis. In addition, its abnormal expression during fibrogenesis is a major feature of fibrotic liver, kidney, and lung disease. Recent studies have described several kinase cascades that can link Ras to transcription factor activation (2)(3)(4)(5)(6)(7). A dominant pathway that is often involved utilizes the Ras 3 Raf 3 MEK 1 3 MAPK series.
Many putative MAPK nuclear acceptor proteins have been suggested in the Egr, Ets, Fos, or AP-1 families (2-7). The ␣I(I) collagen gene contains AP-1 sites present in its 5Ј-UTR and 1st intron. These cites could represent the downstream target of the Ras cascade (8 -15).
We evaluated the role of the Ras-Raf-MAPK cascade during collagen gene expression using cultured hepatic stellate cells (HSC), the major collagen-producing effector cell responsible for hepatic fibrogenesis (16 -21). This system utilizes activated early passage cells, which recapitulate many features of the diseased stellate cell in vivo. (16 -21). This model can serve as a paradigm of the enhanced collagen gene expression, which occurs in vivo during liver injury and fibrogenesis (16 -21). Dominant negative inhibitory mutants were used to specifically consider the roles of Ras, Raf, and MAPK. Overexpressing activated constructs were avoided. Oncogenic forms of ras and raf are not involved in hepatic fibrogenesis. In addition, these constructs may abnormally stimulate a pathway with little physiologic relevance (4 -7, 20, 21).
It was found that blockage of Ras or Raf activity led to an increase in collagen gene expression. This is consistent with the Ras overexpression studies previously mentioned. Surprisingly, however, blockage of MAPK activity decreased collagen gene expression. When both Raf and MAPK activity were simultaneously blocked, collagen gene expression was decreased. These data suggest a branch point between Ras 3 Raf and MAPK. A MAPK-stimulatory cascade is balanced with a Ras 3 Raf-inhibitory cascade. The two separate cascades mapped to two distinct regions of the 5Ј-UTR, unrelated to AP-1 domains. The MAPK cascade involved the ubiquitous Sp-1 and NF-1 transcription factors in the proximal Ϫ100 bp domain. The Ras 3 Raf cascade utilized a more upstream Ϫ1680 bp domain. This latter domain involves a novel 60-kDa DNA-binding protein, which is selectively produced by activated stellate cells in culture and following activation in vivo.

EXPERIMENTAL PROCEDURES
Cell Culture-HSC were isolated from Sprague-Dawley male rats and subcultured by previously described methods (16,17). Experimental manipulations were performed with cells at passages 2-6.
Transfection Studies-Stellate cells were transfected using the Lipo-fectAMINE reagent, and cell extract handling, extraction, quantitation, and CAT measurements were performed as described previously (22,23). Following transfection, the cells were maintained in serum (10% fetal calf, 10% calf serum) for 48 h prior to CAT analysis. The plasmids used for transfection included the intcolCAT plasmids (Ϫ3.6/1.6 or Ϫ1.7/1.6), which contain either Ϫ3.6 or Ϫ1.7 kb of the 5Ј-UTR of the rat ␣I(I) collagen linked to the 1st exon and 1.6 kb of the 1st intron and the CAT reporter gene (9). The Ϫ1.3/1.6 intcolCAT plasmid and the Ϫ0.4/ 1.6 intcolCAT plasmid were derivatives of the Ϫ3.6/1.6 intcolCAT plasmid produced by digestion with NheI/TthIIII and NheI/MfeI restriction endonucleases, respectively. Ends were then blunted and ligated by T 4 DNA ligase. The pdel1.3-0.4/1.6 plasmid was a derivative of the Ϫ1.7/ 1.6 plasmid, which was produced by digestion with TthIIII/MfeI. This created a deletion from Ϫ1284 bp to Ϫ392 bp, and then the ends were blunted and ligated with T 4 DNA ligase. The Ϫ3.6siteTAE/1.6 plasmid was created by overlap extension and contained the mutated TGF␤ activation element (TAE(A)) described below (24). All mutated plasmids were sequenced to confirm the positions of the mutations. All transfections contained equivalent amounts of plasmid DNA (1.5 g) by using empty vector pMNC plasmid. Cotransfections utilized either pMNC or CMV-driven dominant negative raf (301-1 plasmid), dominant negative ras (N-17), dominant negative or wild type MAPK (pCMVp41(Ala-54/ 55) MAPK , G3BCAT (Sp-1 CAT reporter), or BCAT-1 (Sp-1 CAT "control" reporter) plasmids (2-7). Transfection efficacy was monitored by paral- lel transfection with a ␤-galactosidase plasmid, as described previously (22,23). Each data point represents the mean of two sets of six pooled transfections. Variation between sets was Ͻ15%, unless otherwise shown. Results are representative of at least three independent experiments.
To evaluate the requirement for TAE binding in vivo, transfections were performed as above with varying concentrations of doublestranded oligonucleotides (sites Ϫ1625 to Ϫ1615 bp in the 5Ј-UTR of the ␣I(I) collagen gene). The following were the sequences of the oligonucleotides used. Similar wild type and mutated TAEs have been shown previously to selectively disrupt TGF␤ stimulation of collagen promoter activity (25). In addition, TAE protein binding was abolished when mTAE(A) or mTAE(B) was used (25).
Gel Retardation Assays-Nuclear extracts (10 g/lane) from HSCs were incubated with radiolabeled double-stranded TAE oliogonucleotides (Life Technologies, Inc.) and electrophoresed on a native gel, as described (25). Competition binding assays were performed with unlabeled double-stranded oligonucleotides (TAE or mutated TAE).
Southwestern Blotting-Nuclear extracts (50 g) were obtained from either culture-activated stellate cells (HSC) or fresh quiescent HSCs from normal rats. These were compared to stellate cell extracts from activated cells obtained from rats that had been pretreated with carbon tetrachloride (CCl 4 ) (0.5 ml ϩ 0.5 ml of mineral oil given intraperitoneally) 48 or 72 h prior to sacrifice. The extracts were resolved by SDS-PAGE, transblotted to a nitrocellulose membrane, and probed with the radiolabeled double-stranded TAE described above, using standard Southwestern blotting techniques (25).

RESULTS AND DISCUSSION
Dominant Negative raf Versus Dominant Negative MAPK: Differential Regulation of Collagen Gene Expression-Stellate cells cotransfected with a collagen reporter gene and either dominant negative raf plasmid or dominant negative MAPK plasmid yielded divergent changes in gene expression (Fig. 1A). Dominant negative raf transfection resulted in a 5-fold increase in reporter expression. Surprisingly, when dominant negative MAPK was transfected, a 3-fold decrease in reporter expression was found. Previous HSC studies used these same dominant negative plasmids to demonstrate raf's and MAPK's respective roles in insulin growth factor and 1,25-dihydroxyvitamin D 3 nuclear signaling (23). By varying the amount of input DNA, the optimal dominant negative raf or MAPK plasmid concentration was determined (Fig. 1, B and C). Based on previous studies with these same plasmids, the optimal plasmid concentrations should result in sufficient amounts of the mutated kinase proteins to bind the upstream cascade proteins (either Ras or MEK, respectively) (2-7). However, the precise mechanism associated with the dominant negative plasmid effect is unclear. The dominant negative MAPK suppression of reporter expression was observed in the absence (Fig. 1C) or presence of dominant negative raf (Fig. 1D). This reduction in colCAT expression by dominant negative MAPK (1 (base line) versus 0.35 (dominant negative MAPK) (CAT units/mg of protein)) contrasts markedly with the increase in colCAT expression (to 5.5 CAT units/mg of protein) obtained by transfecting comparable amounts of the wild type MAPK plasmid (data not shown). In other studies, dominant negative ras transfection caused a 3.3-fold increase in reporter expression (data not shown), which mimics the dominant negative raf effect. These findings imply that the suppressive pathway is likely to involve a Ras 3 Raf link (2-7). Collectively, these results suggest that Raf activation leads to two distinct effects on type I collagen promoter activity: (i) a suppressive effect via a MAPK-independent pathway and (ii) a stimulatory effect via a MAPK-dependent pathway. The latter effect is also likely to involve Raf-parallel pathways, which utilize other members of the enlarging MAPK kinase kinase family (2,7). These studies further imply that there is an additional branch point between Raf and MAPK. The Ras-Raf-MAPK cascade could help regulate the expression of the major disease-related collagen gene during fibrogenesis. Accumulated evidence suggests that the diseased stellate cell initially responds to mitogenic stimuli (e.g. platelet-derived growth factor), which would be expected to activate Ras 3 Raf. The increase in stellate cell collagen gene

Ras/Raf and MAPK Regulate Collagen Gene Expression 11040
expression occurs at a later time point. The fibrogenic stimuli responsible for this later increase are incompletely understood, but transforming growth factor ␤ (TGF␤) is a likely mediator (17)(18)(19)(20)(21). The TGF␤ kinase cascade involves several unique kinases, but their role in collagen gene expression is unknown (26,27). Previous work suggests TGF␤ may block Ras activation and therefore decrease Raf activation (28). During fibrogenesis, the stellate cell may be initially exposed to activated Raf and MAPK (in a platelet-derived growth factor-dominated stage). At the later TGF␤-dominated stages, activated Raf levels may decrease. This would be predicted to lead to an increase in collagen gene expression. The relative duration of Raf versus MAPK activation or the amount of nuclear activated MAPK could then determine the ultimate extent of collagen gene activation. In view of the development of selective MEK inhibitors, these observations may have therapeutic importance (29).
Dominant Negative raf Versus Dominant Negative MAPK Utilize Different DNA Response Elements-To identify the DNA response elements which are sensitive to the dominant negative (dn) MAPK versus dominant negative (dn) raf effects, truncated reporter constructs were substituted for the parent plasmid. The dn-MAPK-sensitive region of the collagen gene was found to be independent of the 1st intron, which contains AP-1 binding sites (Fig. 2) (12-14). In addition, most of the 5Ј-UTR appears to be dispensable. Recent studies have suggested that the basal promoter activity lies within the most proximal 200 base pairs, which contain consensus Sp-1 binding sites (within a region termed footprint 2) and consensus NF-1 binding sites (within a region termed footprint 1) (8, 30). Sitedirected mutagenesis of the NF-1 site in footprint 1 abolished the response to dn-MAPK (Fig. 2). In addition, cotransfection experiments found that dn-MAPK caused a 50% reduction in a Sp-1-driven CAT reporter versus no effect on a control Sp-1 reporter (data not shown). Therefore, the MAPK-sensitive region appears to involve both NF-1 and Sp-1 sites. Future studies will be needed to determine if this is a direct effect on the phosphorylation state of these transcription factors or an indirect effect involving other MAPK-sensitive nuclear factors.
Using the same approach, the dn-raf-sensitive region was localized to a different upstream region (Fig. 3A). Sequential deletion analysis revealed that the 1st intron is not required, but a region between Ϫ1.7 kb and Ϫ1.3 kb is needed. When this region was placed adjacent to the Ϫ400 bp region of the 5Ј-UTR, the previously unresponsive Ϫ400 bp-containing plasmid (Ϫ0.4/1.6) regained its sensitivity to dn-raf stimulation (8,30). Previous studies in other cell systems have suggested that this Ϫ1.7 to Ϫ1.3 kb region contains significant DNA binding activity at the Ϫ1.62 kb region, termed the TAE, a region previously shown to be required for TGF␤ stimulation of the type I  Fig. 1. The various colCAT reporters are labeled and schematically depicted on the left. The relative amount of colCAT expression is displayed on the right (Ⅺ, pMNC; f, dn-MAPK), with the absolute amount of colCAT expression in the presence of the pMNC plasmid given an arbitrary value ϭ 1 for each individual reporter. The parent intcolCAT reporter contains Ϫ3.6 kb of 5Ј-UTR region upstream of the 1st exon (depicted as a black box), which is serially linked to 1.6 kb of the 1st intron, the SV40 splice acceptor (depicted as a gray box), and then the translation start site and the chloramphenicol acetyltransferase gene (depicted as a striped rectangle). The deletions (del) made in the 1st intron are shown as empty rectangles. The mutation of the NF-1 site of footprint 1 (codon substitutions GG 3 TT at codon Ϫ96 and Ϫ97)) is depicted as a dotted oval in its respective plasmid (Ϫ3.6(FP-1mut)/ 1.6). When this plasmid was used in cotransfections with pMNC, the absolute amount of CAT/mg protein was 4 -5-fold reduced versus the Ϫ3.6/1.6 intcolCAT reporter. The Ϫ3.6siteTAE/1.6 plasmid contained the mutated TAE region (depicted as a hatched oval) (mTAE(A)) in situ.
FIG . 3. Mapping of the dominant negative raf response region. A, dominant negative raf-induced stimulation requires the Ϫ1.7 to Ϫ1.3 kb upstream region of the 5Ј-UTR. HSCs were cotransfected with col-CAT reporter plasmids as described in Fig. 2. The pdel1.3-0.4/1.6 plasmid contained the Ϫ0.4/1.6 plasmid ligated to the Ϫ1.7 to Ϫ1.3 kb region of the 5Ј-UTR. The Ϫ3.6siteTAE/1.6 plasmid contained the mutated TAE region (depicted as a hatched oval) (mTAE(A)) in situ, as in Fig. 2. The relative amount of colCAT expression is displayed on the right (Ⅺ, pMNC; f, dn-raf). B, dominant negative raf-induced stimulation is blocked by excess TAE. HSCs were cotransfected with dominant negative raf and the pdel1.3-0.4/1.6 plasmid Ϯ 1 or 2 g of doublestranded TAE and maintained in serum-containing media for 48 h prior to CAT analysis, as described previously (22,23). Data are expressed (mean Ϯ standard deviation; n ϭ 3) in terms of the amount of dominant negative raf stimulation. An arbitrary value ϭ 1.0 with transfection with wild type TAE is equivalent to no effect with the dominant negative raf plasmid. Parallel cotransfections utilized 1 or 2 g of mutated TAE(A) or TAE(B). Similar results were obtained when the Ϫ3.6/1.6 plasmid was used (data not shown). Ⅺ, TAE; f, mTAE(A); d, mTAE(B). collagen promoter (25). This effect was attributed to a novel 30-kDa protein identified by Southwestern blotting (25). Since preliminary studies suggested that the TAE region was responsible for most of the DNA binding activity of the Ϫ1.7 to Ϫ1.3 kb region in HSCs, further studies concentrated on the TAE region. Site-directed mutagenesis of the TAE region markedly reduced the dn-raf effect (see Ϫ3.6siteTAE/1.6 reporter, Fig.  3A). In contrast, the dn-MAPK effect was preserved when the same TAE mutated reporter (see Ϫ3.6siteTAE/1.6 reporter, Fig. 2) was used. To confirm that TAE binding was required in vivo for the dn-raf effect, HSCs were cotransfected with the pdel1.3-0.4/1.6 reporter and double-stranded TAE in a competition assay (Fig. 3B). When 1 or 2 g of TAE was used, the dn-raf effect was abolished. A lesser blocking effect was seen with 0.5 g of TAE (data not shown). To control for nonspecific binding, two mutants of TAE were used. These mutants (A and B) contain codon substitution in the wild type TAE that eliminate TAE-DNA binding in vitro (see below). When either TAEmut ((A) or (B)) was substituted for wild type TAE, the dn-raf stimulatory effect on colCAT expression was preserved. TAEmut(A) partially suppressed the dn-raf effect but it was much less effective than wild type TAE. This suggests that TAEmut(A) retains some binding activity in vivo that is eliminated in TAEmut(B). The more stringent conditions of the in vitro gel retardation assay (see below), however, suggest that the majority of TAE binding is lost in TAEmut(A). These results collectively demonstrate that TAE binding is required for dnraf stimulation of colCAT expression, and this involves the Ϫ1628 to Ϫ1615 bp domain.

TAE Binding Is Selective for in Vitro and in Vivo Activated
Stellate Cells-TAE characterization studies revealed that HSCs contain a single retarded band when nuclear extracts are incubated with radiolabeled TAE in a gel retardation assay (Fig. 4A). Binding specificity was confirmed by competition binding assays with increasing concentrations of unlabeled TAE but not by similar amounts of unlabeled mutated TAEs (mTAE(A) or mTAE(B)). Southwestern blotting demonstrated a single 60-kDa binding activity (p60) in cultured HSCs (Fig.  4B). This binding activity is likely to have relevance during stellate cell activation and fibrogenesis in vivo because an identical binding activity was found in nuclear extracts from stellate cells activated in vivo after a single injection of CCl 4 . Freshly isolated stellate cells have very low levels of collagen gene expression and lack this DNA binding activity (Fig. 4B) (8,20). CCl 4 treatment increases stellate cell proliferation and collagen gene expression during the immediate 48 -72 h postexposure and induces fibrosis/cirrhosis after 6 -12 weeks of chronic exposure (20,21).
Since the TAE binding domain is dissimilar to known transcription factor binding domains, p60 may represent a novel transcription factor. Future studies will need to characterize this protein further.

FIG. 4. Stellate cells bind the TAE.
A, gel retardation assays demonstrate selective binding to the TAE oligonucleotide. Nuclear extracts (10 g/lane) from HSCs were incubated with radiolabeled doublestranded TAE and electrophoresed on a native 6% polyacrylamide gel. Competition binding assays were performed with varying amounts of unlabeled TAE or unlabeled mutated TAE(A) or TAE(B), as indicated. The arrow on the left indicates the specific binding of TAE. B, activated stellate cells contain a 60-kDa protein binding activity for the TAE. Nuclear extracts (50 g) from either culture-activated (C) HSCs or quiescent HSCs (Q) obtained directly from normal rats or rats given carbon tetrachloride 48 or 72 h prior to sacrifice were resolved by SDS-PAGE, transblotted, and probed with radiolabeled doublestranded TAE, using Southwestern blotting techniques. A single 60-kDa protein (p60) was detected in the lanes containing in vitro or in vivo activated stellate cell extracts but not in the quiescent cell extract. Molecular size markers are shown on the left of the gel.