We analyzed the role of oxidative stress on liver collagen gene expression in vivo. Long- and short-term supplementation with the lipophilic antioxidantd-α-tocopherol (40 IU/day for 8 wk or 450 IU for 48 h) to normal C57BL/6 mice selectively decreased liver collagen mRNA by ∼70 and ∼60%, respectively. In transgenic mice, the −0.44 kb of the promoter and the first intron of the human collagen α1(Ι) gene were sufficient to confer responsiveness to d-α-tocopherol. Inhibition of collagen α1(Ι) transactivation in primary cultures of quiescent stellate cells from these transgenic animals byd-α-tocopherol required only −0.44 kb of the 5′ regulatory region. This regulation resembled that of the intact animal followingd-α-tocopherol treatment and indicates that d-α-tocopherol may act directly on stellate cells. Transfection of stellate cells with collagen-LUC chimeric genes allowed localization of an “antioxidant”-responsive element to the −0.22 kb of the 5′ region excluding the first intron. These findings suggest that oxidative stress, independently of confounding variables such as tissue necrosis, inflammation, cell activation, or cell proliferation, modulates in vivo collagen gene expression.
- liver fibrosis
- stellate cells
although overproduction of collagen type I by hepatic stellate cells (lipocytes) is a critical step in the development of liver cirrhosis (15, 19, 20, 35), the regulation of collagen α1(Ι) gene expression remains unclear (12). From several lines of investigation, we have obtained evidence indicating that aldehyde-protein adducts, including products of lipid peroxidation, modulate collagen gene expression (6, 13, 25) and may be a link between injury and fibrosis (3, 24, 28).
Coculture experiments of hepatocytes and stellate cells treated with carbon tetrachloride (a hepatocyte, but not stellate cell, toxin) indicate that hepatocytes could exert a paracrine stimulation of both lipid peroxidation and collagen gene expression on stellate cells (3). In support of a role of reactive aldehydes on collagen gene expression, both acetaldehyde (6) and malondialdehyde (13) increase collagen α1(Ι) gene transcription, and this effect can be blocked by scavengers of reducing equivalents (13), which are required for the formation of aldehyde-protein adducts. Acetaldehyde, a product of ethanol oxidation, can also induce lipid peroxidation (43). In agreement with these findings, stellate cell collagen gene expression is stimulated by acetaldehyde (38) and by 4-hydroxynonenal (40), another product of lipid peroxidation.
Whether oxidative stress plays a role in the modulation of collagen gene expression in vivo is unknown. The assessment of this important issue is complicated by the confounding variables of tissue necrosis and inflammation associated with tissue injury. These confounding variables could be avoided by analysis of constitutive collagen gene expression. Because lipid peroxidation also occurs in normal tissues in vivo (2, 10, 21, 25, 32), we investigated whether products of lipid peroxidation might modulate basal liver collagen gene expression. In this study, we show thatd-α-tocopherol supplementation results in a decrease of collagen α1(Ι) gene expression in the liver of normal mice and in cultured hepatic stellate cells.
Three-week-old female C57BL/6 mice received either ad-α-tocopherol-supplemented diet (8 IU/g of foodstuff) or a control (Purina Chow) diet for 8 wk. Transgenic mice received intraperitoneal injections of eitherd-α-tocopherol (100 mg, 150 IU) or equal volumes of mineral oil (200 μl) at 0, 24, and 48 h and were killed 6 h after the last injection.
Expression of chimeric collagen α1(Ι) human growth hormone reporter genes in the liver.
Transgenic mouse lines expressing human growth hormone (hGH) transgenes were as described by Bornstein and co-workers (34, 44). We used a transgenic line containing portions of the human α1(Ι) collagen gene, −2300COLΔI (bases −2300 to +1607, with deletion of most of the first intron) and −440COL (bases −440 to +1607, which include the first intron and first exon). The −440COL line has been previously described (26, 44) and contains 2 copies of the transgene, whereas the −2300COLΔI line contains 10–12 copies (26). The cis-regulatory region of the collagen α1(Ι) gene responsive tod-α-tocopherol was characterized in transgenic mice expressing hGH under the direction of −2300 or −440 bp of the human collagen α1(Ι) without or with the first intron, respectively (44). At least three animals were used for each group in all experiments.
Primary hepatic stellate cells from transgenic mice were isolated essentially as described previously (26). Briefly, after the livers were excised and washed in Hanks’ balanced salt solution (HBSS), they were minced and incubated at 37°C for 30 min with constant shaking with 0.5% Pronase (Boehringer Mannheim), 0.05% collagenase B (Boehringer Mannheim), and 10 μg/ml DNase (United States Biochemical) in HBSS without Ca2+. This digest was filtered through gauze and pelleted at 450g for 10 min and subsequently rinsed four times with HBSS containing 10 μg/ml DNase. Primary hepatic stellate cells from rats were isolated as described previously (3, 33). Stellate cells were purified by a single-step density Nycodenz gradient (Accurate Chemical & Scientific, Westbury, NY), as described previously (3, 26, 33). Cells were cultured on EHS (Matrigel) and used after 6 days of culture. In some experiments, stellate cells growing on an EHS matrix were activated with FeSO4 (50 μM)-ascorbate (200 μM), which induces oxidative stress, as we previously described (33). Cells were cultured under an atmosphere of 5% CO2 and 95% air in DMEM containing 10% FCS. Stellate cells were identified by their typical autofluorescence at 328 nm (excitation wavelength), staining of lipid droplets by oil red, and immunohistochemistry with a monoclonal antibody against desmin (3). In addition, the activation of stellate cells was determined by the expression of α-smooth muscle actin, judged by immunofluorescence using specific antibodies against α-smooth muscle actin (Sigma) (33).
Determination of collagen α1(Ι) mRNA.
RNA was isolated from liver and quantitated by a sensitive RNase protection assay, as described previously (7-9, 16, 26), using the riboprobes for 18S RNA (Ambion), exon 5 of the hGH gene, and mouse collagen α1(Ι) (26, 27, 44). Collagen α1(Ι) mRNA and hGH mRNA values were measured simultaneously with the internal standard 18S RNA as described by us previously (7-9, 16, 26, 27).
All the results are expressed as means ± SE. Student’st-test was used to evaluate the differences of the means between groups, acceptingP < 0.05 as significant.
We first analyzed the expression of the collagen α1(Ι) gene in the liver of C57BL/6 mice that received a diet supplemented with the lipophilic antioxidantd-α-tocopherol (8 IU/g of foodstuff, ∼40 IU/day; n = 13) (8). After 8 wk, the weight of animals receiving a control diet (Purina Chow without the d-α-tocopherol supplementation; n = 7) was similar to that of the d-α-tocopherol group (20 ± 1 vs. 21 ± 1 g; not significant). The effects ofd-α-tocopherol supplementation on liver collagen α1(Ι) were analyzed by an RNase protection assay, after purification of poly(A)+ RNA from total liver RNA (5). The steady-state pool of collagen mRNA in the liver was consistently and markedly inhibited (∼70%) by dietary supplementation withd-α-tocopherol (Fig.1, A andB).
Because d-α-tocopherol inhibits collagen α1(Ι) gene transcription in cultured human fibroblasts within 72 h (25), we analyzed whether short-term supplementation withd-α-tocopherol would also effectively inhibit liver collagen gene expression.d-α-Tocopherol (150 IU ip) was administered at 0, 24, and 48 h to C57BL/6 mice. The endogenous collagen α1(Ι) steady-state pool in the liver was decreased by ∼60% in animals treated withd-α-tocopherol compared with control animals (receiving equal volumes of mineral oil vehicle) (Fig.1 C). Confounding variables present during stellate cell activation in vivo and in culture (e.g., cytokine production, cell proliferation) preclude a definitive characterization of the “oxidative stress”-responsive region (3, 26, 33). Therefore, to characterize the oxidative stress-responsivecis-regulatory region within the collagen α1(Ι) gene, we performed experiments with normal transgenic mice expressing the hGH under the direction of regulatory regions of the human collagen α1(Ι) gene (Fig.2 A) (44). Both the hGH transgene mRNA and the endogenous collagen α1(Ι) mRNA were detected by a sensitive and specific RNase protection assay (26). The expression of the −2300COLΔI-hGH transgene and the −440COL-hGH transgene was markedly decreased in the liver, in parallel to endogenous collagen (data not shown) after treatment withd-α-tocopherol (150 IU ip at 0, 24, and 48 h) (Fig. 2 B). In normal liver, a 5′ regulatory region that included only the −0.44-kb region was sufficient to direct the expression of the hGH. In the same animals,d-α-tocopherol also inhibited collagen α1(Ι) gene expression in tendon, a tissue with high collagen production, by ∼60% (K. S. Lee, unpublished observations), suggesting that the bioavailability ofd-α-tocopherol and its therapeutic efficacy, after only a 48-h treatment, are probably adequate for modulating collagen gene expression in many tissues.
Because the effects ofd-α-tocopherol on liver collagen gene expression could be exerted directly or indirectly on stellate cells, we analyzed the modulation of collagen gene expression by d-α-tocopherol in freshly isolated hepatic stellate cells. Stellate cells were obtained from −2300COLΔI-hGH and −440COL-hGH transgenic animals as discussed in methods. The stellate cell populations were >95% pure, as determined by autofluorescence at 328 nm, conferred by retinoids. Contaminant cells were hepatocytes and other sinusoidal cells (3). To avoid the confounding variables of cell activation and proliferation, which are associated with enhanced oxidative stress (33), stellate cells were cultured on an EHS matrix. Under this culture condition, stellate cells displayed for at least 6 days a quiescent phenotype similar to that of stellate cells in normal liver, as described previously (26, 33). The expression of the endogenous collagen α1(Ι) mRNA was inhibited by the antioxidantsd-α-tocopherol and butylated hydroxytoluene (Fig.3 A). Also, expression of hGH in stellate cells bearing the −2300COLΔI-hGH (data not shown) or the −440COL-hGH was markedly decreased from basal values by treatment with 50 μMd-α-tocopherol during a 6-day culture period (Fig. 3 B). Thus transactivation of collagen α1(Ι) gene expression in primary cultures of quiescent stellate cells required only −0.44 kb of the 5′ regulatory region and resembled the regulation of these cells in the intact animal followingd-α-tocopherol treatment. Complementary information about the role of lipid peroxidation on collagen gene expression was obtained by treating control quiescent stellate cells, cultured on EHS matrix, with ascorbic acid (200 μM)-FeSO4 (50 μM), a free radical-generating system (13), for 6 days. This treatment increased ∼10-fold the mRNA steady-state pool of collagen α1(Ι) (data not shown). In addition, we have previously shown that Fe2+-ascorbate induces hepatic stellate cell activation (33). Taken together, these studies indicate that basal oxidative stress modulates hepatic stellate cell collagen α1(Ι) gene expression in vivo and in culture and that the −0.44-kb region of the collagen α1(Ι) gene contains the antioxidant-responsive element(s).
The promoter of the collagen α1(Ι) gene is highly conserved among different species, including humans, mice, and rats (14, 34, 42). Therefore, additional characterization of thecis-regulatory region within the collagen 5′ flanking sequences was obtained by transfecting aLUC chimeric reporter gene driven by a −220-bp segment of the mouse collagen α1(Ι) gene into activated (cultured in 20% serum) or quiescent (cultured in 0.5% serum) primary rat stellate cells. Using a transfection-enhancing reagent and lipofectamine, we achieved a high-efficiency transfection for stellate cells, as discussed previously (27, 29). To obtain optimal reporter expression, cells were harvested 48 h after transfection. Expression of the LUC reporter containing −220 bp of the 5′ flanking region of the mouse α1(Ι) collagen gene (without the first intron) was inhibited byd-α-tocopherol irrespective of whether the cells were activated (20% serum) or quiescent (0.5% serum) (Fig. 3 C). The transfection efficiency, as determined by a pLUC vector, was comparable in control and d-α-tocopherol-treated cells (12–16 U/μg DNA), indicating thatd-α-tocopherol did not decrease collagen-chimeric reporter gene expression spuriously by inhibiting the transfection of this gene.
We have previously suggested that lipid peroxidation plays a critical role in tissue fibrogenesis by stimulating collagen gene transcription (3, 13, 24, 25). Moreover, we have shown that activation of cultured hepatic stellate cells by transforming growth factor-α and collagen type I is mediated by oxidative stress (33). Here, we present evidence strongly supporting the hypothesis that antioxidants, in the absence of necrosis or inflammation, inhibit collagen α1(Ι) gene expression in the liver of normal animals.d-α-Tocopherol also inhibited collagen α1(Ι) gene expression in cultured stellate cells, independently of the confounding variables of activation stimuli.
The effects of d-α-tocopherol were observed after an 8-wk supplementation of ∼40 IU/day or after only a 48-h treatment of 450 IU. These results suggest the feasibility of controlling collagen gene expression in the liver during chronic or acute induction of fibrogenesis. Furthermore, after a 48-h treatment,d-α-tocopherol also inhibited collagen gene expression in tendon, a tissue with high collagen production. Altogether, these studies suggest that the bioavailability of d-α-tocopherol is probably adequate for many tissues and thatd-α-tocopherol inhibits collagen gene expression in tissues with low and high rates of collagen production. These findings could be of great relevance not only for the treatment of excessive fibrogenesis in the liver but also in other tissues such as lung, kidney, skin, pericardium, pleura, and peritoneum.
The precise molecular mechanisms by which oxidative stress regulates collagen α1(Ι) expression in stellate cells are unknown. However, oxidative stress plays an essential role, through the induction of c-myb and nuclear factor-κB, on stellate cell activation, including induction of α-smooth muscle gene expression (33). Whether similar mechanisms also affect collagen α1(Ι) expression in stellate cells remains to be determined. Reactive aldehydes are formed as a result of the oxidative breakdown of polyunsaturated fatty acids, and these aldehydes may form covalent links with various amino acid residues of proteins (4, 43). The functions of various proteins are altered by the formation of aldehyde-protein covalent bonds in vitro (4).
In this study, we also show that induction of free radicals with ascorbic acid-FeSO4 increases collagen gene expression in cultured quiescent stellate cells. Malondialdehyde- and 4-hydroxynonenal-protein adducts have been demonstrated on the induction of ascorbic acid-FeSO4 in cultured fibroblasts, and their formation is inhibited withd-α-tocopherol (13). Presumably, by inhibiting lipid peroxidation and thus the formation of reactive aldehydes such as malondialdehyde and 4-hydroxynonenal,d-α-tocopherol inhibits adduct formation. Presently, no other biologically important function has been ascribed to d-α-tocopherol distinct from its role as an antioxidant. In this context, malondialdehyde- and 4-hydroxynonenal-protein adducts have been found in animals with iron overload (24, 28) and CCl4-induced hepatotoxicity (3,26) as well as in patients with genetic hemochromatosis (30) and chronic viral hepatitis (31).
It is also important to note that aldehydes may form adducts with virtually all cellular elements, including DNA (4, 43). Because endogenous malondialdehyde-deoxyguanosine adducts have been detected in apparently normal human livers (11), it is conceivable that aldehydes may stimulate collagen gene expression by a direct interaction with DNAcis-acting regulatory elements. Although expression of the collagen α1(Ι) gene in the liver during hepatocellular injury and lipid peroxidation induced by CCl4 requires only the presence of an upstream −0.44-kb region (26), it was uncertain whether this transcriptional activation is mediated by oxidative stress or other confounding factors. To characterize the regulatory region of the collagen α1(Ι) gene responsive to oxidative stress, we analyzed the expression of reporter chimeric genes in transgenic mice following treatment with d-α-tocopherol. Here, we report thatd-α-tocopherol inhibits the expression of the −440COL-hGH as well as of the −2300COLΔI-hGH transgene in the absence of liver necrosis or inflammation, strongly suggesting that the antioxidant effect is exerted on the −0.44-kb flanking region of the collagen α1(Ι) gene.
Because in the intact animald-α-tocopherol could modulate liver collagen α1(Ι) gene expression indirectly through paracrine mechanisms, we tested whether the samecis-regulatory region of the collagen α1(Ι) gene contains the “oxidant”-responsive element(s) in primary stellate cell cultures. In primary cultures of quiescent stellate cells (to avoid the confounding variables of cell activation and cell proliferation) bearing the −440COL-hGH transgene,d-α-tocopherol inhibited the expression of the hGH reporter to a degree similar to that observed in the −440COL-hGH transgenic mice. These studies indicate thatd-α-tocopherol can directly modulate stellate cell collagen expression through the −440-bp segment of the 5′ flanking region and/or the first intron. Further characterization of the sequences responsive tod-α-tocopherol was obtained by transfection into stellate cells of a chimericLUC reporter gene. A regulatory region of the collagen α1(Ι) gene containing the −220/+110-bp segment in the absence of the first intron was sufficient for the inhibition of LUC reporter expression by d-α-tocopherol in transfected primary stellate cells.
It was not known whether the mechanisms that modulate constitutive collagen expression in cultured cells also occur in vivo. Cells are usually exposed to a higher oxygen tension in vitro (1) than in vivo (22), and this could induce high levels of basal lipid peroxidation, which in turn may lead to a high constitutive collagen gene expression. However, under the conventional incubation conditions (95% air, 5% CO2), the level of lipid peroxidation found in cultured fibroblasts or myocardial cells (280 ± 30 pmol/mg protein) (17, 25) is comparable with that in normal skin (2, 32) and other tissues in vivo (10, 21, 25, 39). These findings suggest that the degree of lipid peroxidation is comparable in normal cultured cells and normal tissues in vivo. More importantly, basal oxidative stress appears to modulate constitutive collagen gene expression of quiescent hepatic stellate cells both in culture and in the normal liver.
Another important issue is whether the constitutive collagen gene expression is spuriously high in cultured stellate cells or other cells. Several studies seem to indicate that collagen gene expression is quantitatively and qualitatively similar in vitro (6, 13, 18, 23,41) and in vivo (45). Although absolute rates of collagen production are difficult to measure, the data available suggest that the values are equivalent (1–3 μmol proline ⋅ mg protein−1 ⋅ h−1) in normal cultured fibroblasts (18, 23) and in normal skin in vivo (36,37). Also, the steady-state level of collagen α1(Ι) mRNA by solution hybridization is in the same range in stellate cells activated in culture or in vivo (26).
We herein demonstrate that antioxidants, independently of other confounding variables such as tissue necrosis, inflammation, cell activation, or cell proliferation, modulate hepatic collagen gene expression. Whether oxidative stress plays a role in the modulation of constitutive liver collagen gene expression in humans is unknown. Given the findings presented here, and due to the low toxicity ofd-α-tocopherol and other antioxidants, the modulation of collagen gene expression by antioxidants could now be assessed in normal individuals and eventually in patients with active fibrogenesis in the liver or in other tissues.
We are indebted to Dr. P. Bornstein (University of Washington, Seattle, WA) for providing the founder transgenic mice for lines −2300COLΔI and −440COL. We thank D. Walker, K. Pak, and A. Nesterova for technical assistance and L. Masse for the preparation of this manuscript.
Address for reprint requests: M. Buck, Dept. of Medicine, Univ. of California, San Diego, VAMC, 9–111D, 3350 La Jolla Village Dr., San Diego, CA 92161.
This study was supported by National Institutes of Health Grants DK-38652, DK-46971, and GM-47165 and by grants from the Dept. of Veterans Affairs. K. S. Lee was supported by a grant from Yonsei University College of Medicine (Seoul, South Korea).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
- Copyright © 1998 the American Physiological Society