Potential conflict of interest: Nothing to report.
Hepatic stellate cell (HSC) activation is a pivotal step in the pathogenesis of liver fibrosis. The clarification of this transdifferentiation process is therefore important for the development of effective therapies for fibrosis. We analyzed the effect of a histone deacetylase inhibitor, valproic acid (VPA), on mouse HSC transdifferentiation in vitro and in vivo. The exposure of freshly isolated mouse HSCs to 2.5 mM VPA led to increased histone H4 acetylation and inhibited cell proliferation. Expression of stellate cell activation markers analyzed by quantitative polymerase chain reaction and western blotting revealed that treatment with VPA inhibited the induction of activation markers such as Acta2, Lox, Spp1, and Myh11. Treatment of mice with VPA decreased collagen deposition and in vivo activation of stellate cells in the livers of CCl4-treated mice. Class I histone deacetylase silencing through RNA interference in mouse HSCs only partially mimicked treatment with VPA. Conclusion: Chronic administration of VPA results in a marked decrease in stellate cell activation both in vitro and in vivo. We hypothesize that the VPA effect results partially from class I histone deacetylase inhibition, but that also non-histone deacetylase class I VPA targets are involved in the stellate cell activation process. (HEPATOLOGY 2010.)
Hepatic stellate cell (HSC) activation is an initial event in liver fibrosis. During chronic liver injury, HSCs change from vitamin A storing quiescent cells to contractile and secretory myofibroblast-like cells. This transdifferentiation is accompanied by several new phenotypic characteristics, such as enhanced cell migration and adhesion, expression of α-smooth muscle actin (α-SMA), increased proliferation and contractility, loss of retinoid storing capacity and, most importantly, acquisition of fibrogenic capacity. Once contraction and extracellular matrix (ECM) protein secretion become excessive this can lead to impaired organ function.1 This activation process of HSCs is closely reproduced when freshly isolated HSCs are cultured on plastic dishes.2 Gene expression studies have shown that bile duct ligation and CCl4-activated and culture-activated HSCs display an almost identical pattern of up-regulated and down-regulated genes.3 Among these genes is Acta2 (encoding α-SMA protein), the most widely used marker for HSC activation.4
Although liver fibrosis has been studied extensively, drugs to prevent and treat fibrosis are only partially effective.5 For many patients with end stage liver disease, liver transplantation is the only available option. Therefore, studying the underlying mechanisms of HSC activation is an important step toward identification of molecular targets and the development of more effective therapies.4 Alterations in expression of several transcription factors such as JunD, FoxF1, FoxO1, peroxisome proliferator-activated receptor γ, KLF6, and Lhx2 have been associated with the HSC activation process. However, the exact regulation of this event is unknown.6 An important process in transcriptional regulation is the modification of histones, of which the complex regulation can be linked to activation as well as to repression of gene expression. Functionally, histone modifications have the potential to influence several biological processes including differentiation and transdifferentiation.7, 8 Recent studies have shown the importance of epigenetic regulation underlying the transdifferentiation of HSCs in vitro. Mann et al.9 have demonstrated that treatment of cultured rat HSCs with a DNA methylation inhibitor, 5-aza-2-deoxycytidine, prevents activation of the cells. Additionally, our laboratory has identified the histone deacetylase inhibitor (HDI) trichostatin A (TSA) as a potent inhibitor of HSC activation. TSA inhibited synthesis of procollagen type I, procollagen type III, and α-SMA filament formation and HSC proliferation.10–12
Acetylation by histone acetyltransferases often takes place on N-terminal tails of histone proteins and is associated with activation of transcription. Histone deacetylases (HDACs) catalyze the removal of acetyl groups from histone proteins, thereby inducing a positive charge on the lysine side chains of histones H3 and H4 and preventing the access of transcriptional complexes to DNA. Generally, HDAC activity is linked to transcriptional repression. Four families of HDACs have been characterized since the first identification of HDACs in 1996: the class I, II, and IV HDACs and the class III or SIRT family. Recently, a rapidly growing number of nonhistone proteins have been found to be targets for HDACs.13 Over the past few years, more attention has been drawn to HDACs for two main reasons: first, the relationship between HDACs and several diseases, including cancer, has been confirmed; second, many HDIs are used in clinical and preclinical research as anticancer agents and show satisfying effects.14
In the present study, we show that chronic administration of valproic acid (VPA), a more selective class I HDI when compared with TSA,15, 16 results in a marked decrease in stellate cell activation in vitro and in vivo and significant reduction in septa formation and fibrogenesis in vivo. We hypothesize that the VPA effect partially results from class I HDAC inhibition, but also non-HDAC class I VPA targets are involved in the HSC activation process.
Our institution's guidelines for the care and use of laboratory animals in research were strictly followed. Mouse HSCs were isolated from normal and fibrotic livers. The HSC isolation method for male Balbc mice (25-35 g) was a modification of a previously described method for rat HSCs17 (see Supporting Materials and Methods). For in vivo HSC activation, mice underwent eight intraperitoneal injections over 4 weeks of 50 μL CCl4/100 g body weight in mineral oil (Sigma-Aldrich, St. Louis, MO). Mice used for isolation of in vivo–activated HSCs received four injections over 2 weeks. By using this shorter treatment period, we were still able to isolate HSCs based on their lipid content.3 To study the effect of VPA on in vivo HSC activation, mice received drinking water containing 0.4% VPA twice a week, starting 2 days before the first CCl4 injection.18 The half-life of VPA in serum is on the order of 16 hours,19 and peak serum VPA measurements of 3-70 mg/L are obtained in mice using this method.18
Blood samples were taken from the inferior vena cava, centrifuged at 2,000g for 10 minutes, and stored at −20°C. Serum aspartate aminotransferase (AST) and alanine aminotransferase (ALT) activities were determined at 37°C with an automated analyzer using a standardized test system VITROS 5.1 FS (Ortho Clinical Diagnostics, Beerse, Belgium).
Messenger RNA Analysis.
Total RNA from liver tissue and tissue culture cells was extracted using Trizol (Invitrogen, Eugene, OR) and RNeasy kits, respectively (Qiagen, Hilden, Germany) and reverse-transcribed using the High Capacity cDNA Archive kit (Applied Biosystems Foster City, CA). Gene-specific primers and a Universal Probe Library probe were determined using Probe Finder software of Roche (https://www.roche- applied-science.com/sis/rtpcr/upl/adc.jsp). For real-time polymerase chain reaction (PCR), 2× Maxima Probe qPCR Master Mix was used (Fermentas, St. Leon-Rot, Germany), subjected to quantitative PCR (qPCR) in an ABI 7500 Real Time PCR System, and analyzed using System SDS software (Applied Biosystems), using 18S ribosomal RNA for normalization. The fold change differences were determined using the comparative threshold cycle method.
Cell Proliferation Assay.
Cell proliferation was investigated by measuring active DNA synthesis with the Click-iT EdU Cell Proliferation Assay Kit (Invitrogen); 3,750 cells per cm2 were plated in the presence or absence of 2.5 mM VPA. After 48 hours, EdU labeling was initiated. Another 48 hours later (day 4), cells were formalin-fixed and visualization of the EdU incorporation was obtained according to the manufacturer's instructions.
Cells were formalin-fixed after 4 days of culture in presence or absence of 2.5 mM VPA followed by overnight incubation with primary antibodies (acetyl-Histone H4, 1/250 [Upstate Cell Signaling]; smooth muscle actin, 1/1,000 [Sigma]) and/or an Alexa488 labeled phalloidin (1/1,000 [Sigma]). Antibody binding was visualized using Alexa488/647-labeled antibodies (1/200). Images were taken with an Olympus IX 70 confocal microscope (Olympus Belgium, Aartselaar, Belgium).
Twelve-micrometer frozen sections were cut, air-dried, and fixed with SUSA's fixative for 1 hour and stained for 45 minutes with 0.1% Sirius Red F3BA in a saturated picric acid solution. From each section, nine pictures were made using an Axioskop light microscope (Carl Zeiss, Zaventem, Belgium) and the pictures were recorded using an Axiom digital camera. Red staining was quantified using NIH ImageJ software (http://rsb.info.nih.gov/ij/).
Ten micrograms of protein were analyzed by way of western blot analysis according to standard procedures (Supporting Materials and Methods) using the following antibodies and dilutions: HDAC3 (1/250 [BD Transduction Laboratories], recognizing HDAC1, HDAC2, and HDAC3), HDAC8 (1/250 [Santa Cruz Biotechnology]), and α-SMA (1/10,000 [Sigma]).
Small Interfering RNA Transfection.
All small interfering RNAs (siRNAs) used in this study were siGENOME SMART pools from Thermo Scientific Dharmacon; HDAC1: M-040287-03 HDAC2: M-046158-01 HDAC3: M-043553-01 HDAC8: M-058613-01 (Dharmacon, Lafayette, CO). siRNAs (5 nM) were transfected using HiPerFect Transfection Reagent (Qiagen) according to the manufacturer's instructions. Cells were transfected twice over 5 days and collected 4 days after the final transfection. A nonsilencing siRNA was used as a control.
Comparisons between more than two groups were tested using analysis of variance, followed by Tukey's posttest. Statistical analysis of values for comparison between two groups was performed using a two-tailed Student t test.
VPA Partly Inhibits CCl4-Induced Fibrosis.
In vitro, TSA has been shown to inhibit HSC activation,10, 11 whereas in vivo its inhibitory effect on liver fibrosis has never been reported. Because of TSA's limited use in vivo,20 the influence of the HDI VPA on the mouse model of CCl4-induced liver fibrosis was tested because of its preference toward class I HDACs15, 21 and its documented use in mouse models.18, 22 Mice were treated with CCl4 for 4 weeks with or without VPA in their drinking water. The overall appearance of the mice was normal; the treatment did not influence their behavior, body weight, or liver/body weight ratio. Mice were sacrificed and livers were analyzed for markers of fibrosis (Fig. 1). The overall extent of septa formation in livers stained by Sirius Red was smaller in the VPA-drinking animals compared with control animals (many “chicken wires” in control CCl4-treated mice). Quantification of Sirius Red–stained collagen in images of mouse liver tissue clearly shows that CCl4+VPA-treated animals show less collagen deposition than CCl4-treated animals (Fig. 1A). For CCl4-induced chronic liver injury, the effect of VPA on serological markers for liver fibrosis and liver function (ALT and AST) were determined. Serum ALT and AST levels of the CCl4+VPA-treated group were not influenced significantly when compared with serum levels of CCl4 mice (Fig. 1B). RNA analysis by way of qPCR of the livers showed that VPA cotreatment inhibited the CCl4-induced up-regulation of the classical profibrogenic markers Acta2, proCol-1a1, Timp-1, and Mmp13 (mouse homologue of MMP1) (Fig. 1C).
VPA Treatment Inhibits HSC Activation In Vitro.
To investigate whether the inhibitory effect of VPA on fibrogenesis could be due to an inhibition of HSC activation, we incubated freshly isolated mouse HSCs with increasing concentrations of VPA. We observed a clear difference in morphological appearance of HSCs treated with 2.5 mM VPA (Fig. 2A), so we used this concentration for all in vitro studies. Whereas cells cultured under normal conditions clearly underwent transdifferentiation, the VPA-treated cells did not become myofibroblastic, even after 10 days in culture. When the cells were stained for acetylated histone H4 proteins, we observed a clear increase in acetylated histone H4 in the VPA-treated HSCs when compared with control cells (Fig. 2B). Proliferation, a characteristic of transdifferentiating HSCs, was greatly reduced in the VPA-treated HSCs when compared with control HSCs (Fig. 2C). At the protein level, VPA treatment resulted in an inhibition of the strong up-regulation of α-SMA normally observed during HSC activation in vitro (Fig. 2D). Gene expression levels of several genes known to be regulated during HSC activation in vitro and in vivo3, 23 were analyzed during the same 10-day in vitro culture period using qPCR. The strongest VPA-dependent gene expression changes during HSC activation were observed for Acta2 (α-SMA), Myh11 (smooth muscle myosin), Lox (lysyl oxidase), and Spp1 (secreted phosphoprotein 1, osteopontin), whereas Gfap and Timp1 were not influenced (Fig. 3A).
The capacity of VPA to reverse HSC activation was tested by adding VPA to 7-day-old HSC cultures and analyzing whether the cells retrieved their quiescent properties. VPA treatment of activated HSCs induced strong inhibition of Acta2 messenger RNA (mRNA) levels (Fig. 3B [+VPA at day 7]) and a significant change in α-SMA protein expression (Fig. 3C). Morphologically, HSCs treated with VPA at day 7 showed a more quiescent phenotype accompanied by a decrease of α-SMA fibers when compared with control HSC cultures (Fig. 3D). We washed away the VPA after 7 days of treatment and analyzed whether the HSCs could still transdifferentiate into myofibroblasts. Three days after the removal of VPA, HSCs expressed higher amounts of Acta2 and regained their characteristic myofibroblastic morphology (Fig. 3B,D [−VPA at day 7]).
VPA Treatment Inhibits HSC Activation In Vivo.
Next, the in vivo effect of VPA on genes that were VPA-sensitive in our in vitro experiments was investigated. For this, we analyzed the livers used for the experiments in Fig. 1 in which we show that Acta2 expression is altered by VPA cotreatment (Fig. 1C). RNA analysis by way of qPCR revealed that VPA cotreatment also inhibits the CCl4-induced up-regulation of Lox and Spp1 (Fig. 4A [4-week treatment]). To exclude that the observed effect was due to other cell types than HSCs in the fibrotic liver, we isolated HSCs from normal and fibrotic mice with or without VPA in their drinking water. As described by De Minicis et al.,3 we were able to isolate in vivo–activated HSCs from 2-week CCl4-treated mice. Under these conditions, we observed a CCl4-induced up-regulation of Acta2, Lox, and Spp1 in total liver RNA that is reduced by VPA cotreatment (Fig. 4B). Analyzing freshly isolated HSCs from 2-week CCl4-treated mice showed higher levels of Acta2, Lox, and Spp1 when compared with HSCs isolated from control animals. VPA cotreatment inhibited the CCl4-induced up-regulation of Acta2, Lox, and Spp1 (Fig. 4C).
The results described so far were obtained in a prophylactic setup. In order to test the possible therapeutic effect of VPA, we treated mice for 2 weeks with CCl4, followed by 2 weeks of CCl4+VPA cotreatment and compared these with 4-week CCl4-treated mice. Sirius Red stainings showed a significant reduction in collagen deposition when CCl4 treatment was continued in the presence of VPA (Fig. 5A). qPCR analysis for Acta2 in total liver RNA of these mice confirm these results by a reduced expression of Acta2 in the CCl4+VPA mice compared with the CCl4 mice (Fig. 5B). These results suggest that VPA treatment prevents further progression of CCl4-induced fibrosis in mice.
Expression and Function of Class I HDACs During In Vitro HSC Transdifferentiation.
To gain insight in the mechanisms involved in the effect of VPA on HSC transdifferentiation, we determined the expression of class I HDACs during normal HSC differentiation in vitro. Whereas HDAC1 and HDAC2 are easily detected in quiescent (D1) HSCs, their protein expression decreases during stellate cell activation. In contrast, HDAC3 seems to be expressed at constant levels, whereas HDAC8 is induced upon HSC transdifferentiation (Fig. 6A). Because VPA has been shown to not only inhibit HDAC activity but also lead to proteasomal degradation of HDAC2,24 we determined the impact of VPA treatment on class I HDAC protein expression levels in HSCs. When HSCs are exposed to VPA for 4 or 10 days, the protein levels of class I HDACs were clearly inhibited (Fig. 6B). We then used siRNA mediated knockdown of class I HDACs to evaluate their impact on HSC activation. Fig. 6C shows the knockdown of the class I Hdacs in the HSCs at day 9 of culture. Although this class I Hdac knockdown did not affect Acta2 expression in these cultures, the up-regulation of Lox expression was clearly inhibited (Fig. 6D).
Short-Term Transforming Growth Factor-β Signaling in HSCs Is Not Influenced by VPA In Vitro.
Because a class I Hdac knockdown could not mimic the effect of VPA treatment, we looked for other targets of VPA. Transforming growth factor-β1 (TGF-β1) is an important cytokine in the pathogenesis of liver fibrosis because it up-regulates α-SMA and collagen expression.25 Furthermore, it has been shown that modulation of HDACs by TSA affects TGF-β1 signaling in skin fibroblasts.26 Therefore, we tested the effect of VPA on TGF-β signaling. qPCR analysis revealed that the early TGF-β responders Smad6 and Smad7 were not affected by VPA cotreatment. Tgf-β1 mRNA levels were not influenced by either TGF-β1 or VPA treatment. However, up-regulation of Acta2 and Lox expression by longer TGF-β1 exposure was completely inhibited by VPA (Fig. 7).
After liver injury, HSCs differentiate into myofibroblast-like cells that contribute to tissue repair during wound healing, but severely impair organ function when contraction and ECM protein secretion become excessive.1 The involvement of epigenetic regulation during HSC activation was reported in a recent study by Mann et al.9 Treatment of cultured HSCs with a DNA methylation inhibitor prevented the loss of expression of some antifibrotic proteins, such as peroxisome proliferator-activated receptor γ and IκBα. Ten years ago, Niki and colleagues10, 11 introduced an HDI as a candidate to preserve a quiescent HSC phenotype in vitro; however, the role of individual HDACs was not addressed, because the broad spectrum inhibitor TSA was used to inhibit the in vitro HSC activation. Because of TSA's limited use in vivo20, 27 we set out to test the influence of the more selective class I HDI VPA15, 16 on the mouse model of CCl4-induced liver fibrosis.
Since its introduction into clinical use in 1968, VPA has become one of the most widely prescribed antiepileptic drugs worldwide. Overall, the drug is well tolerated by the majority of patients; however, over the years some mild but manageable side effects have been described. The most common adverse effects of valproate include gastrointestinal disturbances, tremor, and weight gain, which are dose-related and reversible through discontinuation of therapy.28 Some cases of VPA-associated hepatotoxicity have been described, predominantly when used in young children under 2 years of age, but with a rapidly decreasing incidence with increasing age. In adults, this hepatotoxicity is idiosyncratic and, according to some studies, partly due to a carnitine deficiency (L-carnitine supplementation reduces the severity of possible VPA-induced side effects).29
To our knowledge, this is the first time that the potential antifibrotic effect of an HDI has been observed in an in vivo model. In mice, VPA hinders fibrogenesis induced by CCl4as demonstrated by a decreased formation of septa and deposition of lower amounts of interstitial collagens in mice that were simultaneously treated with VPA and CCl4 as compared with CCl4-treated animals (Fig. 1). These observations were made in a prophylactic (Fig. 1) and therapeutic (Fig. 5) setup. In addition, analysis of total liver RNA revealed that VPA could prevent the up-regulation of some typical HSC activation markers such as Acta2, proCol-1a1, and Timp-1. However, ALT and AST levels were not significantly altered by VPA treatment, indicating that VPA does not thwart the hepatotoxic effect of CCl4. We could show that VPA treatment maintained a more quiescent cell morphology of HSCs in culture. Along with the inhibitory effects on cell morphology, VPA also inhibited cell proliferation and mRNA up-regulation of several HSC activation markers involved in different cellular processes: Acta2, Myh11, Lox, and Spp1. This antifibrogenic effect of VPA is not restricted to the liver, because recent studies have shown that VPA can also promote quiescence in pancreatic stellate cells in vitro.30 In this study, VPA inhibited Acta2 expression in pancreatic stellate cells, suggesting that the regulation of Acta2 by HDACs is common between stellate cells from different organs. Smooth muscle actin and smooth muscle myosin are contractile filaments characterizing the activation of HSCs and generating calcium-dependent and calcium-independent contractile forces that contribute to cellular contractility. This contraction of HSCs contributes to increased portal resistance during liver fibrosis.31Lox is an enzyme, responsible for cross-linking of collagens and elastins. In many fibrotic processes, Lox overexpression is followed by an excessive cross-linking of ECM proteins resulting in a lower sensitivity toward degradative enzymes and a disruption of the ECM balance.32 Inhibition of Lox up-regulation thus indicates an inhibition of cross-links in the ECM leading to higher accessibility to matrix degrading enzymes. This has been demonstrated in both mice and rats where a Lox-inhibitor, β-aminopropionitrile, decreased liver stiffness.33, 34 Additionally, four LOX-like proteins (Loxl1-4) have been described.35 Of these, RNA levels of Loxl2 and Loxl3 are up-regulated during HSC activation, and this up-regulation can also be inhibited by VPA treatment (Supporting Fig. 2). All four of the LOX-like proteins have the potential to contribute to extracellular stromal stiffness and fibrosis.36 However, there is evidence that LOXL2 more specifically has a role in migration and epithelial-mesenchymal transition in tumor cells.37, 38 However, the function of individual LOX-like proteins in HSCs is unknown. Studies on rat HSCs have shown that Spp1 is involved in a higher proliferation rate and a higher collagen I expression and migratory capacity of the cells during the activation process in vitro.39 Whereas VPA had a clear effect on Acta2, Lox, and Spp1, it did not affect expression of the TSA-sensitive genes Arp2, Arp3, Addl70, and Gelsolin11 (data not shown). F-actin staining of HSCs in the presence or absence of VPA also demonstrates that actin remodeling in general is not affected by VPA treatment (Supporting Fig. 1).
Removal of VPA led to the onset of classical morphological changes associated with HSC activation, indicating that the inhibitory effects of the drug are reversible (Fig. 3D). The expression of key genes normally up-regulated during in vitro HSC transdifferentiation was also inhibited in vivo when CCl4-treated mice were cotreated with VPA. Stellate cells isolated from mouse livers treated with both CCl4and VPA expressed less Acta2, Lox, and Spp1 when compared with CCl4-treated mice (Fig. 4C). A complete inhibition of HSC activation is not observed, because the expression of several HSC activation markers does not seem to be affected by VPA treatment, indicating that the observed inhibition of liver fibrosis by VPA is most likely due to only a partial inhibition of HSC activation. Whereas it has been reported previously that TSA affects the TGF-β1 signaling in skin fibroblasts,26 we show that VPA treatment does not affect the early events following TGF-β1 stimulation of mouse HSCs (up-regulation of Smad6 and Smad7), whereas some late responses to TGF-β1 stimulation are affected (Lox and Acta2). The observation that Lox expression, but not Acta2 expression, was influenced by knockdown of all class I HDACs suggests that class I HDACs do play a role during HSC activation, but that class I HDACs are not the only VPA targets in HSCs involved in their activation process. Interestingly, VPA treatment of HSCs also leads to reduced class I HDAC protein levels (Fig. 6), suggesting that in addition to the inhibition of their activity, VPA can also influence their steady state protein levels. Thus far, this effect has only been reported for HDAC2.24 Most likely, the lower HDAC8 levels are a consequence of inhibition of HSC activation, because this HDAC is up-regulated during normal culture conditions (Fig. 6A,B). This overall VPA-induced reduction in HDAC protein levels was not due to transcriptional regulation of these HDACs (data not shown). Studies in human neuroblastoma SH-SY5Y cells have shown that VPA can influence wnt signaling through phosphorylation of GSK3β on Ser-9.40 Although there is some controversy about the exact role of wnt signaling in HSCs, different studies have shown that wnt signaling is important for HSC activation.41, 42 When we tested whether the effect of VPA on HSC activation could be partly due to induction of GSK3β phosphorylation we observed that VPA did not increase GSK3β phosphorylation in quiescent nor in activated HSCs (data not shown).
Discovery of VPA-sensitive proteins in HSCs, other than HDACs, will be necessary to give more insight into the mechanism behind the VPA-induced inhibition of HSC activation. Identification of transcriptional repressors that regulate Acta2 or Lox and that are transcriptionally up-regulated by VPA are the most likely candidates. On the other hand, transcriptional activators whose activities are negatively regulated by HDACs are also potential applicants for new therapeutic strategies against liver fibrosis.
We remember Professor Albert Geerts, who passed away during the finalization of this study. We are grateful to him for all his enthusiasm and support, which made the realization of this project possible. This paper is in his honor. We express our warmest thanks to Danielle Blijweert, Jean-Marc Lazou, and Kris Derom for their technical assistance and to Tamara Vanhaecke for critical reading of the manuscript.