Differential modulation of rat hepatic stellate phenotype by natural and synthetic retinoids



Activation of hepatic stellate cells (HSC) is a central event in the pathogenesis of liver fibrosis during chronic liver injury. We examined the expression of retinoic acid (RAR) and retinoid X receptors (RXR) during HSC activation and evaluated the influence of natural and synthetic retinoic acids (RA) on the phenotype of culture-activated HSC. The expression of the major RAR/RXR subtypes and isoforms was analyzed by Northern hybridization. Presence of functional receptor proteins was established by gel shift analysis. Retinoic acids, RAR, and RXR selective agonists and an RAR antagonist were used to evaluate the effects of retinoid signalling on matrix synthesis by Northern blotting and immunoprecipitation, and on cell proliferation by BrdU incorporation. The 9-cisRA and synthetic RXR agonists reduced HSC proliferation and synthesis of collagen I and fibronectin. All-trans RA and RAR agonists both reduced the synthesis of collagen I, collagen III, and fibronectin, but showed a different effect on cell proliferation. Synthetic RAR agonists did not affect HSC proliferation, indicating that ATRA inhibits cell growth independent of its interaction with RARs. In contrast, RAR specific antagonists enhance HSC proliferation and demonstrate that RARs control proliferation in a negative way. In conclusion, natural RAs and synthetic RAR or RXR specific ligands exert differential effects on activated HSC. Our observations may explain prior divergent results obtained following retinoid administration to cultured stellate cells or to animals subjected to fibrogenic stimuli. (HEPATOLOGY 2004;39:97–108.)

Hepatic stellate cells (HSCs) exert specialized functions in normal liver tissue: hepatic extracellular matrix turnover, storage of retinyl esters, secretion of a variety of cytokines and control of the diameter of the sinusoids.1 Acute and chronic liver injury activates HSCs to undergo transition into myofibroblast-like cells that play an important role in inflammation and liver tissue repair.2 Myofibroblastic HSCs display enhanced proliferation and synthesize unbalanced amounts of extracellular matrix proteins, matrix-degrading enzymes, and their inhibitors, resulting in matrix accumulation.3 The changes observed in primary HSC cultures resemble the phenotypical changes during the in vivo activation process. Therefore, cultured HSCs are commonly used as a model to study their in vivo role in hepatic tissue repair and fibrogenesis.

In a healthy liver, the most characteristic ultrastructural feature of HSC is the presence of large cytoplasmatic vitamin A-rich lipid droplets, which contain 70 to 80% of the total liver retinol content. Next to enhanced proliferation and matrix synthesis, activation of cultured HSCs correlates with the depletion of their vitamin A esters and a strong reduction of their retinoic acid (RA) levels.4, 5

RAs are important regulators of cell proliferation and differentiation and binding to 2 distinct families of ligand-activated transcription factors, namely retinoic acid receptors (RARα, β and γ) and retinoid X receptors (RXRα, β and γ).6 The natural ligand for the RARs is all-trans retinoic acid (ATRA) although it is also activated by its stereoisomers 9-cis retinoic acid (9RA) and 13-cis retinoic acid (13RA), whereas the RXRs are activated by 9RA only.7 RAR-selective agonists are clinically used for treatment of cancers, acne, and psoriasis, whereas RXR-agonists show potential for the treatment of hyperglycemia in animal models of type II diabetes.8 RARs regulate the transcription of responsive genes as heterodimers with RXRs. In contrast, RXRs play a central role in nuclear receptor signaling, by either forming homodimers or by acting as obligatory heterodimerisation partners for a variety of nuclear receptors (e.g., RARs, peroxisome proliferator activated receptors, vitamin D Receptors). RXR-dimers bind responsive elements, which are identified as direct repeat motifs of the consensus palindromic sequence, AGGTCA.9

Several studies have shown that supplementation of cultured HSCs with all-trans retinol and/or ATRA prevents morphological transition toward the myofibroblast-like phenotype and decreases collagen type I synthesis and cell proliferation.10, 11 Based on these observations it was postulated that retinoids regulate HSC differentiation through activation of RARs and/or RXRs.12, 13 However, retinol and RAs can be metabolized into several active derivatives and little is known about the stability of natural retinoids in HSCs. In a previous paper, we demonstrated that ATRA and 9RA have differential effects on extracellular matrix synthesis and cell proliferation of activated HSC.14 The 9RA significantly reduced HSC proliferation, whereas ATRA showed no significant antiproliferative effect. Despite evidence for the influence of RAs on HSC transdifferentiation, studies focusing on the expression of RARs and RXRs during HSC activation provided conflicting results. Activation of HSC in culture and during experimentally induced fibrosis results in a reduced expression of RARβ and RXRα.4 In contrast, the levels of RARα, β and RXRα remained unchanged following experimental activation of primary HSC by exposure to Kupffer cell-conditioned medium.5 The purpose of this study was therefore 2-fold. First, to evaluate the expression of the different RAR- and RXR-subtypes during conversion of cultured HSC into their activated phenotype. Second, to investigate whether natural RAs in HSC act through activation of RARs and/or RXRs. Therefore, the influence of synthetic RAR- and RXR-selective agonists8 on the synthesis of extracellular matrix proteins by activated HSC, as well as on their proliferation, was evaluated and compared to the effects of natural RAs.


HSC, hepatic stellate cell; RA, retinoic acid; RAR, retinoic acid receptor; RXR, retinoid X receptor; ATRA, all-trans retinoic acid; 9RA, 9-cis retinoic acid; 13RA, 13-cis retinoic acid, RARE: retinoic acid receptor responsive element; mRNA, messenger RNA; RT-PCR, reverse transcription polymerase chain reaction; ECM, extracellular matrix.

Patients and Methods

Isolation of HSC

Adult male Wistar rats (400–500g) were used in all experiments. Animals were treated according to the guidelines of the Council for International Organizations of Medical Sciences, as required by the Belgian National Fund for Scientific Research. HSCs were isolated from rats by collagenase/pronase digestion, as described previously.14 After isolation, HSCs were plated at a density of 1.5 × 105 cells/ml in 250 ml culture flasks (Falcon, Becton Dickinson, Lincoln Park, NJ) and maintained in Dulbecco's modified Eagle's medium (DMEM) (Gibco, Paisley, Scotland) with 10% fetal calf serum, 100 U/ml penicillin and 100 μg/ml streptomycin. Cell growth, purity, and phenotype were evaluated by phase-contrast microscopy and fluorescence microscopy (AXIOVERT100, Zeiss, West Germany). Digital pictures were taken using a Sony Power-Head camera using KS300 Version2 Kontron electronic software (Zeiss).

CCl4 Treatment of Rats

Adult male Wistar rats fed ad libitum were used in all experiments. To induce fibrosis, 5 rats received repeated intraperitoneal injections of CCl4/paraffin (100 μL/100 g body weight) every 72 hours for 5 weeks. Control rats received paraffin oil only. Rats were sacrificed after 5 weeks, and livers were frozen for immunohistochemistry and Northern blotting.

Retinoic Acid Treatment

HSCs were exposed to the synthetic retinoids (E)-4-[2-(5,5,8,8-tetramethyl-5,6,7,8-tetrahydro-2-naphthalenyl)-1-propenyl] benzoic acid (TTNPB), AGN193109, and AGN194204 (Allergan Inc., Irvine, CA) and to ATRA, 9RA, or 13RA (Sigma, Bornem, Belgium) for the indicated times. Preliminary dose-finding studies were performed using .01, .1, 1, 10, and 100 μM of TTNPB, AGN193109, and AGN194204. No toxic effects were observed using concentrations as high as 100 μM, as evaluated by phase contrast and fluorescence microscopy after Hoechst 33342 (20 μg/ml) and propidium iodide (10 μg/ml) incorporation (Sigma).15 Final experiments were performed using .01, .1, 1, and 10 μM of the synthetic retinoids and 1 μM of the natural RAs. Media were renewed every 24 hours. For immunoprecipitations, cells were kept in the presence of 50 μg/ml sodium ascorbate (Merck, Darmstadt, Germany) and 64 μg/ml β-aminopropionitrile (Sigma). All manipulations of solutions and cell cultures were carried out in subdued light. Control cultures were treated with the same final concentration of vehicle.

Cell Proliferation Assay

HSCs were treated with different agonists or vehicle for 48 hours. During the last 16 hours, the cells were labeled by BrdU incorporation. BrdU proliferation assays (Roche, Brussels, Belgium) were performed as described previously.14

Reverse Transcription Polymerase Chain Reaction and DNA Sequencing

Total RNA from liver tissues and cultured HSC was extracted using Qiaquick extraction columns (Operon Technologies, Alameda, CA). Reverse transcription polymerase chain reaction (RT-PCR) was performed using the PCR Core Kit of Perkin Elmer (Applied Biosystems Division, Foster City, CA). Amplifications were run in a Westburg thermal cycler (Operon Technologies). The identity of the PCR products was confirmed by automatic sequencing and by using the Fasta service of EMBL. Fluorescent dye labeled extension products were produced in a single tube using the ABI PRISM TM Dye terminator cycle sequencing reaction kit and loaded on a ABI PRISM 310 (Applied Biosystems).

Northern Blot Analysis

Total RNA was fractionated in a 1% agarose / 3% paraformaldehyde gel and transferred onto a Hybond membrane (Amersham Biosciences, Little Chalfort, England). Hybridizations were carried out using Quickhyb hybridization solution following the manufacturer instructions (Stratagene, La Jolla, CA). Rat specific probes were produced by RT-PCR and labeled with 32P-deoxycytidine triphosphate (32P-dCTP, 3000 Ci/mmol, ICN Biomedicals, Irvine, CA), using the Rediprime II labeling kit (Amersham Biosciences). Filters were exposed to Kodak Biomax-MS using intensifying screens, and radioactive signals were quantitated using a Bio-Rad G-525 Molecular Imager System (Bio-Rad, Hercules, CA).

DNA Binding Assay

Electrophoretic mobility-shifts (EMSA) were performed by radiolabeling ([γ-32P]ATP, Amersham) double-stranded oligonucleotides using T4 polynucleotide kinase (Invitrogen, Carlsbad, CA). DNA-binding reactions were set up using nuclear extracts derived from 7-day-old HSC and/or RAR and RXR proteins synthesized by in vitro transcription/translation of PSG5-RARα, PSG-RARβ and PSG5-RXRα (kind gift of Dr. P. Chambon, Institut de Genetique et de Biologie Moleculaire et Cellulaire, Illkirch, France) using TNT T7 Quick-coupled reticulocyte lysate system (Promega, Madison, WI). For supershift analysis, polyclonal antibodies against RARα, β, γ (Affinity BioReagents, Golden, CO) or RXRα and β (Santa Cruz Biotechnology, CA) were added. Reaction buffers and DNA-binding conditions were as described previously.16

Metabolic Labeling and Immunoprecipitation

HSCs were exposed for 48 hours to retinoids or the same concentration of vehicle. During the last 24 hours, HSCs were metabolically labeled with 50 μCi/ml 35S-methionine/cysteine (Trans 35S-label, specific activity of 35S-methionine > 1,000 mCi/mmol; ICN Biomedicals, Costa Mesa, CA) in methionine-free DMEM supplemented with 10% fetal calf serum, 100 IU/ml penicillin, 100 μg/ml streptomycin. After labeling, radioactive medium and cell layers were collected and processed for immunoprecipitations as described previously.14, 17 Immunoprecipitation results were visualized by autoradiography and quantitated by Phosphor Imager analysis (Molecular Dynamics, Sunnyvale, CA).


All BrdU proliferation assays were done in triplicate or quadruplicate in 5 independent cultures. Statistical evaluation was performed by calculating the 95% confidence intervals. For immunoprecipitations, the amount of immunoprecipitable protein was calculated per dish and per 106 cells for cell layer and medium. The values per 106 cells for medium and cell lysate were expressed relative to the control culture. Values measured for cell layers and medium were added up to calculate the total effect of retinoids at the protein level. Next, the ratios of treated/control were calculated. The number of observations was 5. To determine significances of the difference between controls and retinoid incubated cells, 95% confidence intervals were calculated. For Northern hybridization, data were first normalized to 18S ribosomal RNA, before ratios of treated/control were calculated (n = 3). Data were then analyzed as for immunoprecipitations.


Expression of RAR and RXR Receptors in Fibrotic Livers

To investigate whether RAR and/or RXR receptors might play a role during HSC activation and liver fibrogenesis, we first analyzed the changes in the expression pattern of the different receptor subtypes in normal and fibrotic rat liver tissues. The development of fibrosis was evaluated by picrosirius staining and by immunocytochemistry for glial fibrillary acidic protein, a well-known marker for HSC. The levels of the different RARs and RXRs were determined by Northern blot analysis in normal and fibrotic livers (Fig. 1). RARγ transcripts were almost undetectable in normal liver, but strongly induced in fibrosis (427% ± 149%, P < .01). In contrast, the expression of RARα (124% ± 31%) and RARβ (90% ± 33%) remained unchanged. Both RXRα and RXRβ transcripts were significantly induced (166% ± 35%, P < .05) and (260% ± 25%, P < .01) respectively, whereas no significant differences for RXRγ messenger RNA (mRNA) (128% ± 48%) were observed.

Figure 1.

Expression of RAR and RXR transcripts in fibrotic liver tissue. (A) Representative Northern blot hybridization (paraffin treated controls, C and paraffin/CCl4 treated rats, CCl4). 10μg total RNA was loaded per lane. Black arrows indicate the position of 18S and 28S RNA. (B) Quantitative assessment of the expression of RARα, β, γ and RXRα, β, γ mRNA transcripts in fibrotic liver tissue compared to control liver.

Expression of RAR/RXR Receptor Subtypes During HSC Activation

One of the goals of this study was to determine which RAR and RXR subtypes were expressed by freshly isolated HSC (day 0) and to evaluate their regulation during activation of HSC under culture conditions (days 3, 7, 14, 21, and 28). The expression of the different receptor subtypes in HSC was compared to their expression in total liver tissue. Freshly isolated HSC expressed RARα, β, γ and RXRα and β (Fig. 2A). In contrast to an earlier report,4 we were unable to demonstrate RXRγ in freshly isolated and culture activated HSC. In contrast to total liver tissue, freshly isolated HSC showed strong signals for RARα, γ and RXRβ. Culture activation of HSC resulted in a gradual reduction of RARα, β, γ and RXRα. The levels of RARβ and RARγ were strongly reduced between days 0 and 3 after isolation and gradually decreased at later times. The expression of RXRβ mRNA remained high up to 21 days, thereafter, its expression dropped considerably. To correlate retinoid receptor expression with the phenotypical transdifferentiation of the HSC, cells were examined by phase contrast microscopy every 24 hours for 28 days. Enhanced proliferation was observed between days 3 and 14, whereas no proliferation was observed at later time points.16 After the second passage (day 14), individual cells could be followed through culture, and cultures did not reach confluency. Cells contained 1 or more nuclei, did not divide and were enlarged as demonstrated by the scale bar in Fig. 2B. The size and number of vitamin A-rich lipid droplets was reduced. Autofluorescent retinol-rich droplets were detectable until day 28 of culture.

Figure 2.

Expression of RARs and RXRs in culture activated HSC. (A) Northern blot analysis, showing the expression of RARα, β, γ and RXRα, β, γ. Gene expression was analyzed on days 3, 7, 14, 21, and 28 in 4 independent HSC cultures. TL = total liver. (B) Phenotypical changes in cultured HSC evaluated by phase contrast microscopy. Bar scale = 100 μM.

RAR and RXR Isoform Expression

The complexity associated with the expression patterns of the different RAR and RXR subtypes is increased further by the existence of multiple isoforms generated by differential promoter usage and alternative splicing.18–23 Isoforms are well conserved across species and display a tissue- and stage-specific expression pattern. By RT-PCR, we investigated which of the main RAR/RXR-isoforms were expressed by isolated HSC. No differences in isoform-expression were found between freshly isolated and cultured HSC. HSCs were shown to express RARα1, α2; RARβ2, β4; RARγ1, γ2; RXRα1, α2; and RXRβ1 and β2. Although stellate cells show strong signals for RARγ (vide infra), we were unable to demonstrate significant expression of the RARγ3 isoform, which is the dominant RARγ isoform in mouse liver tissue.20 Weak signals for RARγ3 were found in RNA from whole rat liver. Since only the mouse cDNA sequences were available for RARα, β, γ, we determined their exact rat sequences. The obtained sequences were deposited in the EMBL databank (RARα1, accn°: RNAJ2940; RARα2, accn°: RNAJ2941; RARβ2/β4, accn°: RNAJ2942; RARγ1, accn°: RAN223083).

Presence of an Aberrant RXRα Transcript

By RT-PCR analysis on HSC total RNA, we found 2 amplicons for RXRα (Fig. 3A). The smaller amplicon was sequenced and appeared to be a novel RXRα mRNA transcript. The newly identified RXRα transcript lacked exons 6 and 7 (270 bp), which encode a large part of the ligand-binding domain. Sequence analysis showed that the RXRα ΔE6/E7 variant encoded an in frame splicing, thereby lacking the α-helices H3, H4, H5, and H6 and β-sheets s1 and s2 (Fig. 3B). Helix H7 is only partially absent. All these secondary structures form an essential part of the ligand binding pocket of RXRα, with a hydrophobic cavity formed by H5, s1, s2, and H7 together with the C-terminal part of H10 and N-terminal part of H11 (Fig. 3C).24 This cavity corresponds to the RXR ligand binding pocket. The ligand dependent transactivation domain and tertramerization domain in helix 11–12 remain unchanged. Aberrant forms of nuclear receptors, which act as dominant negative repressors of their normal counterparts, have been found in patients with a variety of endocrine diseases and in cell lines nonresponsive to ligands. An almost identical pattern of alternative splicing and intron-retention in exons 6 and 7 as found for RXRα ΔE6/E7 was previously found in the RXRβ gene and led to aberrant RXRβ-transcripts with dominant negative functions.25 Moreover, the RAR receptors express truncated variants, lacking part of their ligand-binding domain.26 These aberrant RARs act as repressors of normal RARs in a response element and cell-type specific manner.26 Further characterization is needed to establish the functional role, if any, of this novel RXRα ΔE6/E7 transcript. The relative abundance of aberrant nuclear receptors with dominant negative functions suggests that RXRα ΔE6/E7 may be important in regulating the function of RXRα in HSC.

Figure 3.

Presence of an aberrant RXRα transcript. (A) RT-PCR result showing the amplicon of the expected size (546 bp) and the presence of a smaller transcript (276 bp). (B) The RXRα ΔE6/E7 variant transcript shows an in frame splicing, lacking α-helixes H3, H4, H5, H6 and β-sheets S1 and S2 (secondary structures underlined). Primer pair indicated in bold. (C) Schematic representation of the secondary structures (α-helixes and β-sheets) of the RXRα ligand-binding domain and the RXRα ΔE6/E7 variant.24

Functionality of RAR:RXR-Dimers in Activated HSC

To assess whether the reduced amounts of RARα, β and γ remain functionally active, we evaluated the ability of nuclear proteins isolated from 7-day-old HSC to bind to a typical RAR-responsive element (RARE)27 (Fig. 4). Two retarded bands of comparable size were observed. Supershift analysis using RAR and RXR subtype specific antibodies identified the presence of different RAR:RXR-DNA complexes. The upper protein-DNA complex was quantitatively shifted after addition of RARγ antibodies and the formation of this complex was blocked by administration of RXRα antibodies. In contrast, inclusion of RARβ and RXRβ antibodies abolished the formation of the lower protein/DNA complex, whereas the signal of the upper band had become stronger. This may reflect competition for binding to the oligonucleotide between different RAR:RXR dimers and other nuclear receptors, such as PPAR:RXR, VDR:RXR, COUPs, HNF-4, present in the nuclear extracts. No interference with DNA binding was observed using RARα-specific antibody ies. The lack of supershifted bands using anti-RARβ, RXRα, RXRβ antibodies might be due to interference with heterodimerisation or DNA binding or due to instability of the supershifted complex. Non-immune sera did not affect RAR:RXR binding. The specificity of the protein/DNA complexes was further supported by comparison with the DNA-shift observed after heterodimerisation of in vitro-translated RAR and RXR proteins and binding to labeled RARE.

Figure 4.

Electrophoretic mobility assays using oligonucleotides corresponding to the consensus RARE of the CRBP-I promoter. Nuclear extracts obtained from 7-day-old HSCs were preincubated with different RAR/RXR specific antibodies (SS = supershift) and then incubated with labeled oligonucleotide.

Phenotypical Effects of Synthetic and Natural Retinoids

To define the respective roles of RARs and RXRs, HSCs were exposed to a specific RAR agonist (TTNPB) and antagonist (AGN193109) and to a RXR-specific agonist (AGN194204). We evaluated whether these synthetic ligands modulated proliferation and extracellular matrix synthesis of HSC.

BrdU incorporation analysis was performed to determine the effect of synthetic (.01, .1, 1, and 10 μM) and natural (1μM) RAR and RXR ligands on HSC proliferation (Fig. 5). Three day-old and 7-day-old HSCs were exposed to different retinoids (TTNPB, AGN193109, AGN194204, ATRA, 9RA, and 13RA) for 48 hours. During the last 16 hours of culture, the cells were labeled with BrdU. TTNPB, a RAR agonist, showed no effect on HSC proliferation, whereas the RXR agonist AGN194204 significantly inhibited HSC proliferation. Proliferation of HSC was significantly induced following exposure to the RAR-specific antagonist, AGN193109. In contrast to synthetic RAR agonists, proliferation of HSC was significantly reduced after exposure to ATRA, 9RA, and 13RA. The strongest antiproliferative effect was found for 13RA.

Figure 5.

Effect of retinoids on HSC proliferation. (A) Effect of different concentrations of a RAR specific agonist (TTNPB) and antagonist (AGN193109) on BrdU-incorporation by 7-day-old HSCs. (B) Antiproliferative effect of the RXR specific agonists (AGN194204) on 7-day-old HSCs. (C) Synthetic RAR and RXR ligands (1μM) have identical effects on the proliferation of 3-day- and 7-day-old HSCs. (D) Antiproliferative effect of natural retinoids (ATRA, 9RA and 13RA, 1μM) on HSCs. Data are expressed as mean ± SD (n = 5).

Effect of Retinoids on ECM Synthesis

Seven-day old HSCs were exposed to different synthetic RAR and RXR agonists (.01, .1, 1, and 10 μM) for 48 hours before analysis. During the last 24 hours, media were renewed and cells metabolically labeled with 35S-methionine/cysteine. Conditioned media and cell layers were collected, processed separately, and immunoprecipitated with specific antibodies to collagen type I, collagen type III, and fibronectin.14, 17 Figure 6A shows a typical immunoprecipitation result for de novo synthesized collagen type I, collagen type III, and fibronectin excreted into the culture media, demonstrating the dose responsive effect of exposure to TTNPB. Figure 6B summarizes the average quantitative immunoprecipitation data of 1 and 10 μM of the different retinoids (n = 5). Quantitation clearly showed that TTNPB and AGN194204 suppressed collagen I and fibronectin synthesis significantly. Collagen III synthesis was significantly reduced by TTNPB (both at 1 and 10μM) and by 10 μM AGN194204. No significant effects were observed at lower concentrations (.01 and .1 μM) of these retinoids. AGN193109, at all concentrations investigated, did not significantly alter the de novo synthesis of the investigated matrix proteins. We, therefore, investigated whether these retinoids acted at the transcriptional level by lowering specific mRNA levels. Figure 7 shows representative autoradiographs of the Northern hybridization experiments (day 7, n = 3). The RAR agonist TTNPB significantly lowered mRNA levels for all investigated proteins at 10 μM (procollagen α1(I), –63.2 ± 12.3%; procollagen α1(III), –25.4 ± 8.9%; fibronectin, –39.2% ± 17.7%, P < .05), whereas lower concentrations were ineffective. In contrast, the RAR antagonist AGN193109 did not alter the expression of the investigated genes. The RXR agonist, AGN194204, reduced procollagen α1(I) (–39.3 ± 11.0% and –41,7 ± 14.2%) and fibronectin (–31.6 ± 5.3% and –37.8 ± 15.3%) at 1 and 10μM, whereas collagen α1(III) mRNA levels were not altered. No significant effects were observed at lower concentrations (.01 and .1 μM).

Figure 6.

Effect of synthetic retinoids on matrix synthesis. 7-day-old HSCs were exposed to different concentrations of TTNPB (RAR agonist), AGN193109 (RAR antagonist) or AGN194204 (RXR agonist) for 48 hours. During the last 24 hours, cells were metabolically labeled with 35S-methionine/cysteine. Culture media and cell layers were harvested and immunoprecipitated separately with specific antibodies to procollagen type I, procollagen type III, and fibronectin. (A) A representative immunoprecipitation result showing the concentration dependent effect of TTNPB (.01 up to 10 μM) on the deposition of de novo synthesized procollagen type I, procollagen type III, and fibronectin in the culture media. (B) Quantitative assessment of de novo synthesized procollagen type I, procollagen type III, and fibronectin, immunoprecipitated from culture media (gray) and cell layers (dark gray). The ratios of retinoid-treated versus control (vehicle-treated) were calculated. Data are expressed as mean ± SD (n = 5).

Figure 7.

Northern hybridization analysis of connective tissue protein transcripts (collagen type I, CI; collagen type III, CIII; fibronectin, Fib). 7-day-old HSCs were exposed to 1 and 10 μM TTNPB, AGN193109, or AGN194204 (n = 3).

Effect of Retinoids on the Expression of RAR/RXRs

Activated HSCs (day 7, n = 3) were exposed to synthetic (.01, .1, 1, and 10μM) RAR/RXR ligands and RAR/RXR receptor subtype expression was analyzed by Northern blot analysis (Fig. 8). One μM of the RAR agonist TTNPB significantly induced the expression of RARα (+52.1 ± 17.5%) and RARβ (+52.6 ± 11.4%), while RARγ and RXRα and RXRβ mRNAs remained unchanged. Exposure to 10 and 1μM of the RAR antagonist AGN193109 resulted in significantly reduced RARα (–34.4 ± 9.9%) and RARβ (–34.1 ± 7.8%) receptor levels, while RARγ, RXRα and β expression was not influenced. Exposure to 10μM of the RXR specific agonist AGN194204 reduced significantly the expression of RARα mRNA (–57.5 ± 8.7%), whereas the mRNA levels of the other RARs and RXRs were not affected. No significant changes in receptor transcript levels were observed at lower retinoid concentrations.

Figure 8.

Northern blot analysis showing the effect of synthetic retinoids on the expression of RAR/RXRs. Seven-day-old HSCs were exposed to 1 and 10 μM TTNPB, AGN193109, or AGN194204 (n = 3).


The literature concerning the influence of retinoids on HSC remains controversial. Incubation of culture activated HSC with retinol and/or ATRA has been shown to result in a recovery of their intracellular vitamin A reserves,28 a reduction of collagen I synthesis13 and reduced proliferation.10, 11, 29, 30 13RA was shown to prevent rat HSC transformation and to reduce endothelin-B receptor expression.31 HSC isolated from rats pretreated with retinyl palmitate synthesized less ECM and proliferated slower than control cells.32 In contrast, administration of RAs augmented collagen α1(I), plasminogen activator, plasmin, and total and active TGF-β in activated HSC.12, 33–36 Under the influence of TGF-β, ECM synthesis may be up regulated and proliferation inhibited. In contrast, retinoids exert direct effects on ECM turnover by modulating the production of MMP-1 or TIMP-1.37 Taken together, the literature does not allow for firm conclusions as to the effect of naturally occurring retinoids on HSC. Previously, we demonstrated that ATRA and 9RA had differential effects on 14-day-activated HSC.14 To investigate whether these observations reflected the differential interaction of the stereoisomers with RARs or RXRs and the activation of their responsive genes, we examined in the current study the effects of synthetic RAR- and RXR-selective ligands on the phenotype of cultured HSC.

As described previously, HSC proliferation was inhibited by the natural retinoic acids, 13RA, and 9RA, and to a lesser extent by ATRA. HSC proliferation was also inhibited by RXR-selective agonists. Importantly, synthetic RAR agonists did not affect HSC proliferation. These results indicate that the inhibitory effects of retinoids on HSC proliferation are independent of interaction with RARs. In agreement with this interpretation, electrophoretic mobility assays showed that, although RAR transcript levels are reduced, active RAR:RXR complexes remain present in activated HSC. Moreover, the expression of RARα and RARβ remained inducible by RAR agonists and was inhibited by antagonists. Growth suppression by RAs in human carcinoma cells has been correlated with the activation of RXR:RXR dimers, but not of RAR:RXRs.38 Although our observations on matrix synthesis did not suggest a link between the RAs used and TGFβ,33, 36 retinoids have been shown to inhibit cell growth by inducing TGFβ in many different cell types. This possibility remains to be investigated. Another explanation might be found in the mannose 6-phosphate/insulin-like growth factor II receptor (M6P/IGF2R), which plays a critical role in regulating cell growth, and which has been identified recently as a novel retinoid receptor.39 Binding of latent TGFβ to M6P/IGF2R is essential for its activation, whereas binding of IGF2 to the receptor leads to degradation of this mitogen. Binding of retinoids to M6P/IGF2R was shown to mediate their growth inhibitory effects.40 These antiproliferative effects were not inhibited by the RAR antagonist, AGN193109, nor were they mimicked by a RAR-selective agonist (TTNPB), indicating that growth inhibition was independent of RAR signaling.40 A link between retinoid-M6P/IGF2R binding, TGFβ activation and growth inhibition could help to explain many of the conflicting observations concerning the effects of retinoids on HSC activation and fibrogenesis. We previously demonstrated that the expression of M6P/IGF2R is strongly induced during the process of HSC activation in vitro as well as following CCl4-induced liver injury.41

Previously, it was shown that inhibition of RARα blocked the formation of TGFβ in a RA treated stellate cell line.36 Our results using a synthetic RAR antagonist, however, demonstrate that RARs play an important role in the negative control of HSC proliferation, since exposure to a synthetic RAR antagonist significantly induced cell proliferation. AGN193109 antagonizes RAR-signaling by increasing the interaction between the receptor and its corepressors.42 This finding argues against the view of Okuno et al.,33 who proposed the use of RAR antagonists as potential anti-fibrotic agents. Negative regulation of genes is an important aspect of retinoid signalling. Retinoids have been shown to downregulate the promoter activity of the procollagen α1(I) gene by decreasing the binding of RAR-RXR complexes to the RARE sequence.43 Moreover, ATRA and expression vectors for RARβ and RXRα were shown to suppress the activation of the procollagen α2(I) promoter in HSC.44 However, RAR and RXR ligands can also suppress the collagenase gene.45 In our experiments, synthetic RXR ligands significantly reduced the synthesis of collagen type I and fibronectin, but showed no significant effect on collagen type III expression. As previously shown for ATRA,14 the synthesis of collagens type I, type III, and fibronectin was reduced by exposure to a RAR agonist, but required the administration of relative high concentrations of this compound.

The negative regulation of proliferation by RARs, together with their reduced expression during the first 14 days of culture, was consistent with the short period (day 3 up to 14) that cultured rat HSC actively proliferate and lose most of their retinoid reserves.46, 47 It is tempting to speculate that the reduced expression of RARs with a growth-inhibitory function could be 1 step along the path of stellate cell activation during fibrogenesis. The sustained expression of RXRβ is consistent with the central role of RXRs in nuclear receptor signaling. Recently, we demonstrated that HSC proliferation and activation in vitro and in vivo correlates with a strong induced expression of PPARβ.16 Activation of PPARβ enhances HSC proliferation. PPARs and RARs compete for dimerisation with RXRs on palindromic response elements spaced by 1 bp.48 The relative abundance of RARs and PPARs might thus determine whether HSC proliferation is suppressed or induced.

Compared to total liver tissue, we showed that HSCs are enriched with transcripts for RARα, γ and RXRβ. Whereas RARα and RXRβ are expressed in most tissues, and seem involved in the regulation of general cellular functions, the expression pattern of RARγ in rodents is much more limited.23 This adds up to the finding that the expression of RARγ in the liver seems limited to HSC and suggests a special function of this receptor in the phenotype of the cells. Disruption of the RARγ gene in F9 cells has been associated with loss of the RA-inducible expression of laminin B1 and collagen type IV and resulted in reduced RA metabolism, while loss of RARα was associated with an increased RA metabolism.49 Porcine serum induced fibrosis has been correlated with an enhanced expression of RARα and β,33 whereas the expression of RARβ and RXRα was reduced in a bile duct ligation model of fibrosis.4 In our study, CCl4-injury resulted in a strongly enhanced expression of RARγ and RXRβ, whereas these transcripts were barely detectable in normal liver tissue. Whether our observations reflect increased numbers of activated HSCs or are causally related to retinol-accumulation and metabolism during fibrogenesis remains to be investigated. Recent studies showed that the expression of cellular retinol binding protein-I is enhanced in HSC in fibrotic liver tissues,50 and is accompanied by increased vitamin A-storage in HSC aligning the fibrotic septa.16, 51 This typical accumulation of vitamin A in activated HSC during fibrogenesis might have an important role acting as an antioxidant. Together, these novel findings oppose the view that loss of retinoid reserves might represent a causal factor in the activation of HSC, but once more stress the importance of the HSC-retinoid-metabolism during the pathogenesis of fibrosis. It remains poorly documented whether the observed depletion of vitamin A reserves in culture reflects the actual depletion of retinyl ester-reserves of in vivo activated HSC. Recently, it was demonstrated that CCl4-injury elevates hepatic retinol storage in vitamin A-deficient animals52 and low vitamin A levels and malnutrition were shown to enhance experimentally induced fibrogenesis.53 Administration of retinol, retinyl palmitate, RA, or the acyclic derivative polyprenoic acid before, during, and after CCl4 treatment inhibited the progression to and/or induced regression of hepatic fibrosis.35, 54, 55 In contrast, it is well known that chronic hypervitaminosis A leads to severe liver damage and cirrhosis in humans.56 The liver retinoid levels of rats fed a vitamin A-rich diet were significantly reduced following CCl4-treatment and the development of fibrosis was more pronounced.52 Moreover, vitamin A potentiates the toxic and fibrogenic effects of ethanol in rats.57 9RA, ATRA, and the synthetic analogue E-5166, activated plasminogen and resulted in increased levels of active TGF-β and increased synthesis of collagen in porcine serum induced fibrosis.12, 34 Thus, both surplus and shortage of retinoids may enhance the progression of fibrogenesis. In this regard, it is important to mention that genetic disruption of aryl hydrocarbon receptor expression, a transcription factor that controls RA catabolism, results in the accumulation of retinol and RAs in the liver, increased transglutaminase II levels, activation of TGFβ, and the development of a liver fibrogenic phenotype.58

In summary, we have demonstrated that natural RAs and synthetic RAR or RXR-specific ligands exert differential effects on activated HSC. 9RA and synthetic RXR agonists reduce HSC proliferation and synthesis of procollagen type I and fibronectin. ATRA and RAR agonists both reduce the synthesis of all investigated extracellular matrix proteins, but showed a different effect on cell proliferation. Synthetic RAR agonists did not affect HSC proliferation, indicating that retinoids inhibit proliferation independent of their interaction with RARs. RAR specific antagonists enhance HSC proliferation and demonstrate that RARs control proliferation in a negative way.


The authors thank Jean Marc Lazou and Annick Hagers for their excellent technical assistance.