Hepatic Stellate Cell-derived Delta-like Homolog 1 (DLK1) Protein in Liver Regeneration*

Background: Hepatic stellate cells (HSCs) are activated in liver regeneration, but its significance is unclear. Results: DLK is induced in HSC activation. Its neutralization causes HSC quiescence via de-repression of Pparγ and attenuates liver regeneration. Conclusion: DLK1 activates HSCs via Wnt pathway and epigenetic repression of Pparγ to contribute to liver regeneration. Significance: A novel role of DLK1 in liver regeneration is identified. Hepatic stellate cells (HSCs) undergo myofibroblastic activation in liver fibrosis and regeneration. This phenotypic switch is mechanistically similar to dedifferentiation of adipocytes as such the necdin-Wnt pathway causes epigenetic repression of the master adipogenic gene Pparγ, to activate HSCs. Now we report that delta-like 1 homolog (DLK1) is expressed selectively in HSCs in the adult rodent liver and induced in liver fibrosis and regeneration. Dlk1 knockdown in activated HSCs, causes suppression of necdin and Wnt, epigenetic derepression of Pparγ, and morphologic and functional reversal to quiescent cells. Hepatic Dlk1 expression is induced 40-fold at 24 h after partial hepatectomy (PH) in mice. HSCs and hepatocytes (HCs) isolated from the regenerating liver show Dlk1 induction in both cell types. In HC and HSC co-culture, increased proliferation and Dlk1 expression by HCs from PH are abrogated with anti-DLK1 antibody (Ab). Dlk1 and Wnt10b expression by Sham HCs are increased by co-culture with PH HSCs, and these effects are abolished with anti-DLK Ab. A tail vein injection of anti-DLK1 Ab at 6 h after PH reduces early HC proliferation and liver growth, accompanied by decreased Wnt10b, nonphosphorylated β-catenin, p-β-catenin (Ser-552), cyclins (cyclin D and cyclin A), cyclin-dependent kinases (CDK4, and CDK1/2), p-ERK1/2, and p-AKT. In the mouse developing liver, HSC precursors and HSCs express high levels of Dlk1, concomitant with Dlk1 expression by hepatoblasts. These results suggest novel roles of HSC-derived DLK1 in activating HSCs via epigenetic Pparγ repression and participating in liver regeneration and development in a manner involving the mesenchymal-epithelial interaction.

ing to the most proximal GN box within the promoter (8). Activated canonical Wnt signaling in turn leads to epigenetic repression of Ppar␥ involving the binding of the methyl CpGbinding protein 2 (MeCP2) to the Ppar␥ promoter and increased H3K27dimethylation at the Ppar␥ exon (9), leading to the loss of this critical adipogenic gene expression and consequent HSC cell fate alteration into myofibroblastic cells (8).
Activated HSCs also support liver regeneration via release of mitogens such as hepatocyte growth factor (10), pleiotrophin (11), and epimorphin (12). Liver regeneration induced by partial hepatectomy (PH) is associated with the loss of HSC-associated vitamin A content (13) and up-regulation of the HSC activation markers such as ␣-smooth muscle actin (␣SMA), IL-6, and hepatocyte growth factor (14), suggesting that HSCs undergo activation and support a regenerative response. Regulation of HSC myofibroblastic activation by the neurotrophin receptor p75NTR is actually crucial for supporting liver regeneration (10). p75NTR is selectively expressed by HSCs and induced in myofibroblastic HSCs. In p75NTR-deficient mice, liver injury resulting from plasminogen deficiency is exacerbated due to insufficient hepatocyte regeneration resulting from deficient hepatocyte growth factor release by HSCs despite suppressed collagen and ␣SMA expression (10). Thus, activation of HSCs has both fibrogenic and proregenerative purposes, and the best therapeutic strategy for liver disease should be aiming at inhibiting the former while promoting the latter. Adding to this complexity is the role of activated HSCs in liver oncogenesis which commonly accompanies liver fibrosis and cirrhosis. Activated HSCs produce tumor stroma (15) and promote tumor metastasis (16) and progression (15) where active cross-talk between tumor cells and HSCs is implicated (17).
Delta-like 1 homolog (DLK1/Pref-1) is a single-span transmembrane protein with its extracellular domain composed of six epidermal growth factor (EGF)-like repeats. It belongs to the EGF family of homeotic proteins which includes Notch receptors and their ligands known to regulate cell fate and differentiation. DLK1 regulates several differentiation processes including adipogenesis (18 -21), osteogenesis (22,23), neuronal (24), and neuroendocrine (25) differentiation. It is also expressed in malignancies (26) and embryonic hepatoblasts (27) and promotes neuroblastoma cell stemness and tumorigenicity in a manner responsive to hypoxia (28). An extracellular region can be cleaved by the metalloproteinase ADAM-17 (TNF-␣-converting enzyme, TACE) to release a soluble 50-kDa protein with biological activity (29). Molecular mechanisms by which DLK1 signals to achieve these biological effects are still elusive. DLK1 may interact with Notch to modulate its signaling (30) or IGF-binding protein 1 to modulate IGF local adipogenic action (31). DLK1 is also recently shown to bind fibronectin and trigger integrin-mediated signaling such as FAK-Rac-ERK activation to inhibit adipogenesis (32).
The present study has identified selective expression of Dlk1 by HSCs among adult liver cells and its up-regulation in HSC activation in vitro and in experimental liver fibrosis and regeneration. DLK1 activates HSCs via epigenetic repression of Ppar␥ in a manner dependent on canonical Wnt. Dlk1 expression in HSCs is under the control of positive cross-interactions with other morphogens such as Wnt, necdin, and Shh, and most importantly, Dlk1 up-regulated in liver regeneration after PH supports early hepatocyte proliferation and liver growth via a mechanism which appear to involve Wnt10b.

MATERIALS AND METHODS
Animal Models, Cell Isolation and Culture, BrdU Incorporation-HSCs were isolated from male Wistar rats with or without experimental liver fibrosis or mice following PH as described previously (33) by the Non-Parenchymal Liver Cell Core of the Southern California Research Center for ALPD and Cirrhosis. Liver fibrosis was induced in rats by 10-day cholestasis due to ligation of the common bile duct or by provision of phenobarbital in drinking water (500 mg/liter) plus subcutaneous injection of CCl 4 (1 l/g of body weight) given as a 2-fold dilution with mineral oil twice a week for 3 weeks. Male C57BL/6 mice were subjected to 70% PH, and liver RNA was extracted at 1, 2, 4, 8, and 24 h, as well as every day afterward until the 7th day. HSCs and hepatocytes were isolated 1 day after PH. Mice subjected to 70% PH were also injected via tail vein with anti-DLK antibody or normal rabbit IgG (Abcam) in saline (27 g/100 l per mouse) at 6 h after PH to determine the effects on liver growth and biochemical parameters. BrdU was diluted with PBS and injected intraperitoneally to mice 5 h before sacrifice at 1, 2, and 3 days after PH. Frozen liver sections were made, and BrdU immunohistochemistry staining was performed with the BrdU in Situ Detection kit (BD Pharmingen). The collagen promoter-GFP (Coll-GFP) transgenic mice obtained from Dr. David Brenner's laboratory at University of California San Diego were also used for isolation of liver mesenchymal cells from E13.5 embryonic or adult livers (34). The use of animals for this study was approved by the Institutional Animal Care and Use Committee of the University of Southern California and Department of Veterans Affairs Greater Los Angeles Healthcare System. HSCs were cultured on plastic with low glucose DMEM supplemented with 10% fetal bovine serum (FBS) and antibiotics for 1 day or 7 days for analysis of quiescent or activated HSCs. HSCs from the liver fibrosis models were cultured on plastic in the medium containing 2% FBS and analyzed immediately after overnight culture. Cell morphology was assessed by phase contrast microscopy, intracellular vitamin A content by UV-excited autofluorescence, and intracellular lipid by Oil Red O staining. For promoter analysis via transient transfection, the spontaneously immortalized cell line (BSC) established from experimental cholestatic liver fibrosis (35) or Huh7 hepatoma cell line was used. Kupffer cells were isolated by an essentially identical procedure except for the use of the cells at the arabinogalactan gradient interface of 1.043/1.058 and 1.058/1.075 and subsequent adherence method as described previously (36). Hepatocytes were isolated by the standard method of in situ collagenase digestion of the liver and low speed centrifugation (50 ϫ g, 1 min). Sinusoidal endothelial cells were isolated by magnetic cell sorting using SE-1 antibody as described previously (37). Purity and viability of the cells isolated exceeded 95% for all cell types. NIH3T3L1 cells were treated with the adipocyte differentiation mixture MDI (0.5 mM isobutylmethylxanthine, 1 M dexamethasone, and 1 M insulin) or dimethyl sulfoxide as a solvent and used for analysis.
Fluorescence-activated Cell Sorting (FACS)-GFP ϩ cells were sorted from E13.5 embryos or adult livers of the Coll-GFP mice by FACS as described previously (38). The E13.5 embryonic livers were digested with trypsin-EDTA. The nonparenchymal cells (NPCs) were obtained from the adult livers by the collagenase perfusion method (8). The embryonic liver cells or adult nonparenchymal cells were subjected to FACS using FACS Vantage SE (BD Biosciences) (38). Wild-type mice were used as a negative control.
RNA Extraction and Real Time PCR-Total RNA was extracted from the cells using the RNeasy Mini kit (Qiagen). RNA was reverse-transcribed to cDNA by using a SuperScript III First-strand Synthesis system (Invitrogen) and amplified by 40 cycles using the primers listed below and the SYBR Green  Fig. 6D were the same as described previously (38).
Immunoblot Analysis-HSCs were cultured in a 10-cm dish for 7 days followed by infection with Ad.LacZ.shRNA or Ad.Dlk1.shRNA described below at 100 multiplicity of infection for additional 3 days. The cells were then washed with PBS once, and nuclear and cytosolic proteins were isolated as described previously (3). An equal amount of the nuclear or cytosolic extract (20 g) was separated by SDS-PAGE and electroblotted to nitrocellulose membranes. Antibody against DLK (Abcam), p-AKT, AKT, p-ERK, ERK, p-cMET, or cMET, ␤-catenin, nonphosphorylated ␤-catenin, and phospho-Ser-552-␤-catenin (Cell Signaling Technology), was incubated with a membrane at a concentration of 0.2-2 g/10 ml in TBS (100 mM Tris-HCl, 1.5 M NaCl, pH 7.4) with 5% nonfat milk followed by incubation with a horseradish peroxidase-conjugated secondary antibody (Santa Cruz Biotechnology) at 1 g/10 ml. Proteins were detected by a chemiluminescent method using an ECL kit (Amersham Biosciences).
Construction and Use of Recombinant Adenovirus Vectors-Replication-incompetent recombinant adenovirus expressing necdin shRNA was constructed using the BLOCK-iT Adenoviral RNAi Expression system (Invitrogen) according to the manufacturer's instructions. To select the best shRNA against the rat and mouse Dlk1 gene, we first designed four shRNA oligonucleotides by using the Invitrogen shRNA designer. Of these, at ϩ375 (5Ј-GGACGGGAAATTCTGCGAAAT-3Ј) was shown to be most effective. An additional sequence of CACC was added at the 5Ј end, and AAAA was added to the 5Ј end of the complementary sequence. These two DNA oligonucleotides were annealed to generate dsDNA, which was subsequently cloned into the pENTR/U6 vector using the BLOCK-iT U6 RNAi Entry Vector kit. The U6 RNAi cassette in the pENTR/U6 necdin shRNA vector was transferred to the adenoviral expression plasmid by LR recombination reaction using Gateway LR Clonase II Enzyme Mix and pAd/BLOCK-iT-DEST Gateway Vector kit. Isolated adenoviral expression clones were then digested with PacI to expose the inverted terminal repeats and transfected into 293A cells using Targefect F-2 (Advanced Targeting Systems) for production of a crude adenoviral stock. Large scale amplification of adenoviral vector was performed in 293A cells as described previously (3,4). The titer of the purified virus was determined by the standard plaque-forming assay with 293A cells. An adenovirus expressing ␤-galactosidase shRNA (Ad.LacZ.shRNA) was constructed as a control shRNA vector. Necdin silencing efficiency was tested in day 6 culture-activated rat HSCs with a multiplicity of infection of 50, 100, and 200. Adenovirus expressing GFP, PPAR␥, and a dominant negative mutant of PPAR␥ (gifts from Dr. Krishna K. Chatterjee of the University of Cambridge), or Dkk-1 (a gift from Dr. Calvin Kao, Stanford University), was similarly amplified and purified, and their titers were determined.
Transient Transfection and Reporter Gene Assay-To determine the regulation of DLK1 by necdin, Wnt, or DLK1 itself, the Huh7 human hepatocarcinoma cell line was transiently transfected with the full-length Dlk1 promoter (Ϫ1950/ϩ36)-luciferase construct (a gift from Dr. Sang Hoon Kim, Kyung Hee University, Seoul) by using Targefect F-2 followed 6 h later by infection with Ad.LacZ.shRNA, Ad.necdin.shRNA, or Ad.Dkk1 at 100 multiplicity of infection for 2 days. For Dlk1 promoter co-transfection assays, the cells were co-transfected with different amounts of a necdin (a gift from Dr. Yoshikawa, University of Osaka) or DLK1-HA expression vector, a LEF or dnTCF expression vector (gifts from Dr. Randall T. Moon, University of Washington, Seattle, WA). For testing the effects of Shh inhibition on DLK1 promoter activity, Huh7 cells transiently transfected with the promoter-reporter, were treated with different concentrations of cyclopamine for 2 days. The TOPFLASH promoter-luciferase construct (containing eight wild-type T cell factor (TCF)/lymphoid enhancer-binding factor binding sites, a gift from Dr. Randall T. Moon) was also co-transfected with different amounts of a DLK1 expression vector in Huh7 cells to assess the effects of DLK1 on the canonical Wnt pathway. The cell lysate was collected for determination of both firefly and Renilla luciferase activities using the Dual-Luciferase reporter assay system (Promega), and the results were normalized by Renilla luciferase activity.
Chromatin Immunoprecipitation (ChIP)-For testing whether DLK1 silencing decreases MeCp2 binding to the Ppar␥ promoter, carrier ChIP was performed using Raji cells as the source of carrier chromatin. For this analysis, Raji cells (1.4 ϫ 10 7 ) were added to cultured HSCs (0.2 ϫ 10 6 cells) with or without DLK1 silencing or Wnt3 or Ad.dnPPAR␥ treatment and briefly fixed with 1% formaldehyde on the rotating platform for 10 min at room temperature followed by the addition of glycine to a final concentration of 0.125 M. Scraped HSCs and Raji cells in the medium were spun down and washed with cold PBS with protease inhibitors. After lysis of the cells with SDS buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1) with protease inhibitors, the lysate was sonicated and snap-frozen in aliquots. For ChIP, diluted samples were first precleared with protein G-agarose and then incubated with antibody against MeCP2 (Abcam) at 1 g/l at 4°C overnight followed by precipitation with protein G-agarose. After elution of immunoprecipitated complex, cross-linking was reversed with 5 M NaCl and proteins digested with protease K. Extracted chromatin was subjected to real-time PCR using the primers flanking a segment within the Ppar␥ promoter as described recently (9). Ct values of the samples with nonimmune IgG were subtracted and compared with their respective input Ct values.
Statistical Analysis-All numerical data are expressed as mean Ϯ S.D., and a significance of a difference was determined by a two-tailed t test.

Dlk1 Is Selectively Expressed by HSCs in Adult Rat Liver and
Induced in HSC Activation-We first examined the expression of Dlk1 in four different cell types isolated from male adult rat livers by quantitative PCR and immunoblot analyses. Dlk1 mRNA levels are at least severalfold higher in HSCs than other cell types (Fig. 1A). Immunoblot analysis reveals a distinct ϳ60 kDa band and weaker expression of small molecular species (Ͻ50 kDa) in 1 day (D1) quiescent HSCs, but not in sinusoidal endothelial cells or Kupffer cells (Fig. 1B). Hepatocytes show a faint expression of these proteins which may be attributable to inevitable contamination of a small percentage of HSCs after the conventional isolation technique (39). Expression of these molecular forms of DLK1 due to alternative splicing and posttranslational modifications have been described previously (21,40,41), and that with most abundant expression (ϳ60 kDa) is believed to be the full-length form. HSCs which have undergone "activation" after 7-day culture on plastic (D7), show a clear induction of DLK1 much like that seen in 3T3L1 preadipocytic fibroblasts treated with dimethyl sulfoxide. In contrast, adipocytes generated after the treatment of 3T3L1 cells with the adipocyte differentiation mixture MDI, show attenuated expression (Fig. 1B). Next, we examined the expression of Dlk1 in activated HSCs isolated from rats with experimental liver fibrosis induced by the ligation and scission of the common bile duct or repetitive injection of CCl 4 . In both models, isolated HSCs show up-regulation of Dlk1 mRNA expression along with that of ␣1(I) procollagen with conversely suppressed expression of the HSC differentiation or quiescence gene Ppar␥ (Fig. 1C). These results demonstrate that Dlk1 is selectively expressed in HSCs in the liver and induced in HSC activation in vitro and in vivo.
Dlk Knockdown Reverses Activated HSCs to Quiescent Cells-We next expressed shRNA against Dlk1 via an adenoviral vector in culture-activated HSCs to determine the role of Dlk1 in cell fate regulation. In 48ϳ72 h after infection with the adenovirus expressing the shRNA (Ad.Dlk1.shRNA), expression of the major full-length DLK1 is reduced by 60ϳ70% compared with the cells infected with the control virus expressing shRNA against LacZ (Ad.LacZ.shRNA) ( Fig. 2A). This manipulation reverses the phenotype of activated cells to that of quiescent or differentiated HSCs with an increased content of intracellular vitamin A as assessed by UV-excited vitamin A autofluorescence and increased lipid content detected by Oil Red O staining (Fig. 2B). This morphologic reversal is accompanied by suppressed expression of the genes associated with HSC activation such as necdin (Ncd), ␣1(I) procollagen (Coll), canonical Wnts (Wnt10b and Wnt3a), and Shh but increased expression of the quiescence genes such as Ppar␥ and C/ebp␣ (Fig. 2C). We further tested the regulatory role of DLK1 in canonical Wnt signaling by testing the effects of a DLK1 expression vector on the TOPFLASH promoter activity. For this experiment, we used Huh7 cells because activated HSCs or HSC line already express high levels of DLK1. As shown in Fig. 2D, the TOPFLASH activity is significantly increased by the expression of DLK1. Combined with the results on the effects of Dlk1 knockdown, these data demonstrate the positive regulation of the canonical Wnt pathway by DLK1.
Dlk1 Knockdown Epigenetically Derepresses Ppar␥ in Manner Dependent on Wnt Suppression and PPAR␥ Activity-Activation of HSCs is caused by the loss of Ppar␥ expression, and ectopic PPAR␥ expression in activated HSCs restores their quiescence (3,4). The loss Ppar␥ expression is due to the epigenetic repression mediated by MeCP2 (8,9). MeCP2 orchestrates heterochromatin formation at the Ppar␥ locus by not only binding to the promoter and recruiting HDAC and the co-repressor HP-1␣, but also up-regulating EZH2 and increasing H3K27di-or trimethylation at the 3Ј exons (8,9). As expected from our previously study (8), Wnt3a addition further increases MeCP2 enrichment to the Ppar␥ promoter in culture-activated HSCs (Fig. 2E, left panel, 3rd bar versus 1st bar). Expression of a dominant negative mutant of PPAR␥ by an adenoviral vector (Ad.dnPPAR␥) also increases the enrichment (4th versus 2nd bar) whereas expression of PPAR␥ (Ad.PPAR␥) effectively suppresses it (5th versus 2nd bar), suggesting the importance of intact PPAR␥ activity in preventing MeCP2-mediated epigenetic repression of Ppar␥. Next, we determined whether the restored Ppar␥ expression and HSC phenotypic reversal by Dlk1 knockdown are accompanied by reduced MeCP2 binding to the promoter. Indeed, the Dlk1 knockdown significantly reduces MeCP2 enrichment to the promoter (6th versus 1st bar). Because Dlk1 knockdown suppresses canonical Wnts (Fig. 2C) which mediate Ppar␥ epigenetic repression (8), we asked whether reduced MeCP2 enrichment by the knockdown is abrogated by addition of exogenous Wnt3a. This treatment indeed blocks the reduction in MeCP2 binding and even increases it over the control level (8th versus 6th bar), suggesting that the epigenetic derepression is dependent on Wnt suppression. Next, we examined the effects of Ad.dnPPAR␥ and Ad.PPAR␥ with Dlk1 knockdown. Overexpression of dnPPAR␥ but not wild-type PPAR␥ prevents the reduction in MeCP2 binding achieved by Dlk1 knockdown, suggesting that this derepression effect is also dependent on intact PPAR␥ activity. Parallel analysis for Ppar␥ and ␣1(I) procollagen mRNA levels demonstrate that restored expression of Ppar␥ and suppressed ␣1(I) procollagen expression by Dlk1 knockdown are also blocked by Wnt3a addition or dnPPAR␥ expression (Fig. 2E,  middle and right panels). Ad.PPAR␥ raises Ppar␥ mRNA to the level achieved by Dlk1 knockdown and reduces procollagen expression as achieved by Dlk1 knockdown (the last three bars of Fig. 2E, middle and right panels). To examine a potential link between Wnt suppression and PPAR␥ activity which are both required for the derepression the Ppar␥ gene, we assessed the  MARCH 23, 2012 • VOLUME 287 • NUMBER 13 effects of dnPPAR␥ on the expression of canonical Wnt3a and Wnt10b. As predicted, Dlk knockdown down-regulates both Wnt3a and Wnt10b, but dnPPAR␥ overexpression completely abrogates this effect (Fig. 2F), suggesting that PPAR␥ activity itself down-regulates the canonical Wnts which are inducible by DLK1. In support of this notion, expression of PPAR␥ reduces both Wnt3a and Wnt10b to the level achieved with Dlk shRNA, and a combination of Dlk knockdown and PPAR␥ expression does not have an additive effect, suggesting that PPAR␥ expression which is maximally restored by Dlk knockdown, renders maximal inhibition of Wnt expression. Collec-tively, these results suggest that DLK1 negatively regulates expression of PPAR␥ via epigenetic repression and up-regulates canonical Wnt expression by reducing PPAR␥-mediated inhibition.

DLK1 in Liver Regeneration
Dlk1 under Control of Cross-interactions with Other Morphogens-We next examined how Dlk1 expression in HSCs is regulated by other morphogens such as Wnts, necdin, and Shh shown to be involved in activation of HSCs (6,8,42). In day 7 culture-activated HSCs, Dlk1 mRNA expression is suppressed by inhibition of canonical Wnt signaling with the Wnt co-receptor antagonist Dickkopf-1 (DKK1), necdin shRNA, and the Shh inhibitor cyclopamine (Fig. 3A). Dlk1 promoter activity is up-regulated by expression of TCF and inhibited by expression of dnTCF or the Wnt co-receptor antagonist DKK1 (Fig. 3B) in the activated HSC line BSC, suggesting that the canonical Wnt signaling induces Dlk1 expression via its promoter activation. Dlk1 promoter is similarly activated by necdin and reduced by necdin shRNA (Fig. 3C). Because our earlier work identifies necdin as a positive regulator of Wnt10b transcription (8), we next tested whether necdin-induced Dlk1 promoter activation is dependent on the canonical Wnt pathway. Indeed, necdin-induced promoter activity is completely abrogated by expression of DKK1 (Fig. 3C, right panel). In fact, DKK1 even suppresses the basal promoter activity, suggesting the activity of endogenous necdin in activating the Wnt pathway and Dlk1 promoter. Cyclopamine also suppresses Dlk1 promoter activity (Fig. 3D). Collectively, these results suggest that Dlk1 is under a positive regulation by Wnt, necdin, and Shh. Because Dlk1 knockdown also reduces the expression of Wnt10b, Wnt3a, necdin, and Shh (Fig. 2C), there appear to be positive cross-interactions between DLK1 and other morphogens. In addition, DLK1 also has a self-inductive effect as DLK1 expression induces and DLK1 knockdown reduces the Dlk1 promoter activity (Fig. 3E).
Dlk1 Up-regulation in Liver Regeneration-Compelling evidence suggests that activated HSCs participate in liver regeneration. To test the potential role of HSC-derived DLK1 in liver regeneration, we first examined the expression of Dlk1 mRNA in total RNA extracted from mouse regenerating livers during the first 24 h and a subsequent 7-day period after 70% PH. Dlk1 expression is very low before PH and gradually up-regulated Ͼ40-fold at 24 h, and this induction takes place after inductions of known mitogen and co-mitogen genes such as Hgf, Tnf␣, and Il6 at 1ϳ4 h after PH (Fig. 4A, left). The Dlk message is reduced toward the basal level at day 3-7 (Fig. 4A, right). Induction of necdin and Wnt10b begins to occur at day 2-3 and with much lower magnitudes, and Shh induction is late at day 7 (Fig. 4A). This induction of Dlk1 at 24 h prompted us to examine HSCs and HCs isolated at this time point. HSCs isolated at 24 h after PH show a Ͼ4-fold increase in Dlk1 mRNA level compared with HSCs isolated at 24 h after Sham operation (Fig. 4B). HCs isolated from Sham mice have a very low level of Dlk1 expression, but it increases significantly in PH animals at 24 h (Fig.  4B). These HSCs and HCs were co-cultured immediately after isolation with HSCs on the bottom of the wells and HCs on the insert (Fig. 4C). DNA synthesis of HCs was determined by [ 3 H]thymidine incorporation in the absence or presence of anti-DLK1 antibody in the medium. HCs from PH have a significantly increased proliferation compared with Sham HCs. Addition of anti-DLK1 antibody to the culture medium abolishes this increment of PH HC proliferation but has no effect on Sham HCs (Fig. 4C). In a separate set of the identical experiment, Dlk1 mRNA levels were determined in both HSCs and HCs. Induction of Dlk1 expression in both cell types of PH mice is completely abrogated with the anti-DLK1 antibody (Fig. 4D), suggesting the existence of autocrine and paracrine loops.
Next, we tested the effects of the neutralizing antibody in the PH model in vivo. For this experiment, the antibody or an equal amount of nonimmune IgG (27 g) was injected via tail vein at 6 h after PH, the time point Dlk1 mRNA begins to increase. Prior to sacrificing, BrdU was injected for subsequent assessment of HC DNA synthesis by immunohistochemistry. BrdU incorporation into PH HCs is consistently lower at days 1, 2, and 3 in the animals treated with anti-DLK1 antibody than those with injection of nonimmune IgG (Fig. 4E). The liver growth as assessed by a fraction of liver weight over body weight is also reduced by the neutralizing antibody treatment at day 1 (p Ͻ 0.05) and day 2 (p ϭ 0.05 using a two-tailed t test), but a significant difference is not observed at days 3 and 7 (Fig. 4F). These inhibitory effects on liver regeneration are associated with suppressed expression of Dlk1 (supporting anti-DLK1 antibody sufficiently blocked self-induction), p75Ntr, Wnt10b, and Ptn mRNA but not Tnfa, Il6, and Hgf (Fig. 5A); decreased G 1 (cyclin D) and S (cyclin A) phase cyclines and their partner cyclin-dependent kinases (CDK4 and CDK1/2), as well as suppressed levels of p-ERK1/2 and p-AKT at days 1, 2, and/or 3 as highlighted by boxed blots in Fig. 5B. These results suggest that DLK1 induced in both HSCs and HCs after PH, has an important role in inducing growth factor genes such as Wnt10b, Ptn, or p75Ntr and that neutralization of DLK1 results in diminished growth signals and early liver growth after PH. We became particularly interested in Wnt10b because it induces canonical Wnt signaling which in turn activates many of cell cycle genes shown to be suppressed by anti-DLK1 antibody. To address this possibility, we examined the protein levels of nonphosphorylated (stabilized) ␤-catenin and total ␤-catenin in liver protein extracts of day 2 PH livers. Anti-DLK1 antibody treatment indeed reduces the level of stabilized ␤-catenin by 75% and total ␤-catenin modestly (Fig. 5C). ␤-Catenin is phosphorylated at Ser-552 by activated AKT, and this modification increases its nuclear translocation and activity. Because p-AKT is reduced by the antibody treatment (Fig. 5B), we thought p-␤catenin (Ser-552) levels may also be reduced. Indeed, p-␤catenin (Ser-552) levels are reduced by 55% by the antibody treatment (Fig. 5C). Collectively, these results suggest that the DLK1-Wnt10b-␤-catenin pathway contributes to HC proliferation in liver regeneration and that HC DLK1 is induced by HSC-derived DLK1. To test the latter possibility, we have performed a cross-culture experiment in which HCs from Sham were co-cultured with HSCs from Sham or PH. Both Dlk1 and Wnt10b expression are induced in Sham HCs when co-cultured with PH HSCs as opposed to Sham HSCs, and these inductions are abrogated by addition of anti-DLK1 antibody in culture (Fig. 5D), supporting the notion that HSC-derived DLK1 mediates a mitogenic response in HCs via the DLK1-Wnt10b pathway.
DLK1 Expression in Mouse Embryonic Liver-DLK1 is known to be expressed by hepatoblasts (27), but our findings in the liver regeneration model suggest that DLK1 may also be expressed by HSCs in developing liver. To examine this question, we first performed immunostaining for DLK1 along with desmin and ␣SMA on chicken embryonic livers at E5 when the first lobe is developed. Embryonic HSCs which surround islands or codes of hepatoblasts are stained positively for desmin and ␣SMA (Fig. 6A), supporting the notion that embryonic HSCs have an activated phenotype. These HSCs are also positively stained for DLK1. In addition, a subpopulation of FIGURE 4. A, Expression of Dlk1, cytokines, and growth factors after PH is shown. Note that Dlk1 is gradually induced after induction of Tnf␣, Il6, and Hgf and peaks at 24 h (left panel). Dlk1 induction is followed by milder up-regulation of Wnt10b and necdin at 2ϳ3 days and that of Shh at day 7. B, HCs and HSCs isolated at day 1 after PH were show increases in Dlk1 mRNA levels in both cell types. *, p Ͻ 0.05 compared with the respective cells isolated from Sham-operated mice. C, HCs and HSCs isolated at day 1 after PH were co-cultured to determine DNA synthesis of HC in the absence or presence of anti-DLK1 antibody. Note that increased DNA synthesis by PH HC is abrogated with anti-DLK1 antibody. *, p Ͻ 0.05 compared with Sham HC with nonimmune IgG; †, p Ͻ 0.05 compared with PH HC with nonimmune IgG. D, in an experiment identical to that in C, Dlk1 mRNA levels were determined in both cell types. Note that Dlk1 up-regulation seen in both cell types is completely abrogated by addition of anti-DLK1 antibody. *, p Ͻ 0.05 compared with the non-IgG treatment. E, administration of anti-DLK1 antibody given 6 h after PH significantly attenuating HC DNA synthesis as determined by BrdU incorporation at days 1, 2, and 3 after PH in mice. *, p Ͻ 0.05 compared with nonimmune IgG treatment in PH mice. F, anti-DLK1 antibody treatment described in E reducing liver growth as assessed by the liver weight/ body weight fraction at day 1 and 2 after PH. *, p Ͻ 0.05 compared with IgG treatment. Error bars, S.D. hepatoblasts is diffusely positive for DLK1. In mouse embryos at E9.5, the foregut endoderm and WT1 ϩ septum transversum mesenchyme, from which HSCs originate (43), express DLK1 (Fig. 6B). At E13.5, in addition to albumin ϩ hepatoblasts, ALCAM ϩ mesothelial cells and submesothelial cells, which are recently shown to be the precursor of HSCs (43), express DLK1 at the liver surface (Fig. 6B). Desmin ϩ HSCs and perivascular mesenchymal cells are occasionally positive for DLK1 (Fig. 6B). In E17.5 embryos, DLK1 expression decreases in hepatoblasts (Fig. 6B) as reported previously (27) and persists in ALCAM ϩ mesothelial cells and submesothelial cells (Fig. 6B). In addition, some desmin ϩ HSCs and p75NTR ϩ ␣SMA ϩ perivascular mesenchymal cells clearly express DLK1 in E17.5 livers (Fig. 6B). Next, we isolated liver mesenchymal cells using collagen promoter-GFP (Coll-GFP) transgenic mice from E13.5 embryonic or adult livers by FACS (Fig. 6C). Because embryonic HSCs are devoid of fat/vitamin A and their specific surface marker is yet to be identified, this GFP ϩ cell population, which includes HSCs, submesothelial cells, and perivascular mesenchymal cells, was the best we could study. As shown in Fig. 6D, GFPA ϩ populations from E13.5 and adult livers are enriched with desmin and ␣1(I) procollagen (Coll)-expressing cells. Compared with the E13.5 GFP ϩ cells, the expression level of Dlk1 mRNA decreases significantly in the adult GFP ϩ cells, similar to Lhx2 and Ptn, genes expressed in fetal liver mesenchymal cells (Fig. 6D). Conversely, the expression levels of Gfap and Ppar␥ increase in the adult GFP ϩ cells. These results demonstrate that DLK1 is expressed in HSC progenitor cells and HSCs as well as in hepatoblasts in developing livers in a manner similar to our observations in adult mouse regenerating livers.

DISCUSSION
The findings from the present study demonstrate that DLK1 is selectively expressed in HSCs in the adult liver and up-regulated in HSC activation in culture, experimental liver fibrosis, and regeneration. Dlk1 knockdown reverses activated HSCs to fat-storing quiescent cells via epigenetic derepression of Ppar␥ and restored expression of PPAR␥ in a manner dependent on suppression of canonical Wnts. From the established function of DLK1 as an antiadipogenic mediator (19,21), this finding further enforces the concept that the loss of adipogenic regulation underlies a shift of HSC cell fate to myofibroblastic cells (3)(4)(5)(6)(7)(8)(9). A recent publication reports the causal role of DLK1 in HSC activation and the efficacy of Dlk1 silencing for attenuating experimental fibrosis induced by CCl 4 (44). The findings presented here provide the epigenetic basis for the observed anti-fibrotic efficacy of Dlk1 silencing involving HSC fate regulation. However, activation of HSCs has multidimensional functions in a wound-healing response of the liver including their contribution to liver regeneration. Indeed, the present study identifies DLK1 as a novel, early mediator of liver regeneration in the PH model. Thus, it may not be logically correct to blindly target DLK1 or HSC activation for a therapeutic purpose for liver disease as it may potentially impairs a regenerative response.
In liver regeneration after PH, HCs which normally do not express Dlk1 up-regulate expression of this gene. This is consistent with DLK1 expression by embryonic hepatoblasts (27). Addition of anti-DLK antibody suppresses proliferation and Dlk1 expression by PH HCs co-cultured with HSCs, suggesting a paracrine and/or autocrine mode of action. Because we also utilized the no-contact co-culture model, these results suggest that the cross-talk effects must be mediated via a cleaved soluble form as shown for regulation of adipocyte differentiation (29). Further, our results also demonstrate a self-inductive effect of DLK1 on Dlk1 transcription (Fig. 3E), strengthening the possibility that HSC-derived DLK autoinduces its expression in HSCs while having the same inductive effect on HCs. It is also possible that DLK1 up-regulates other mediators which in turn renders proliferative signals in HCs. DLK1 positively cross-interacts with Wnts, necdin, and Shh (Fig. 3, A-D). After PH in vivo, induction of these morphogens lags behind the early Dlk1 up-regulation in liver regeneration (Fig. 4A), suggesting that they may be downstream effectors of DLK1 in this model. The in vivo neutralization experiment shows significant inhibitory effects of HC DNA synthesis and liver growth during early liver regeneration. This inhibition is concomitant with diminished early regenerative signals such as decreased p-ERK1/2, p-AKT, and cyclins/CDKs. The HC mitogens Wnt10b and Ptn are down-regulated by anti-DLK1 antibody treatment in vivo (Fig. 5A) or ex vivo (Fig. 5D). Reduced levels of stabilized ␤-catenin by the antibody treatment are indicative of suppressed canonical Wnt signaling, which in turn may attenuate growth signaling such as p-ERK and p-AKT. Certainly, some of cell cycle genes such as CyclinD1 are direct targets of canonical Wnts. Further, our results from the cross-culture experiment with Sham HCs and PH HSCs, strongly suggest that HSC-derived DLK1 induces the DLK1-Wnt10b pathway in HC as a mitogenic mechanism. However, other mediators may still be downstream of DLK1 in HCs and HSCs to participate in liver regeneration including Ptn and p75Ntr.
The anti-DLK1 antibody inhibitory effect on liver regeneration was transient and appeared to dissipate in the later time points. This may be due to a transient or incomplete neutralization of DLK1 by a single injection of the antibody at 6 h FIGURE 6. A, chicken embryonic liver at E5 shows expression of desmin, ␣SMA, and DLK1 in HSCs surrounding islands or codes of hepatoblasts as detected by immunohistochemistry. Diffuse staining of DLK1 in a subpopulation of hepatoblasts is also evident. Magnification, ϫ200. B, immunohistochemistry of E9.5, E13.5, and E17.5 mouse embryos. In E9.5 embryos, DLK1 is expressed in both the foregut endoderm (fg) and WT1 ϩ septum transversum mesenchyme (stm). h, heart. In E13.5 embryos, DLK1 is expressed in albumin ϩ hepatoblasts and ALCAM ϩ mesothelial cells (arrowhead) and submesothelial cells (arrows) at the liver surface. Some desmin ϩ HSCs and perivascular mesenchymal cells express DLK1 (arrows). In E17.5 embryos, expression of DLK1 becomes weak in hepatoblasts but persists in ALCAM ϩ mesothelial cells (arrowhead) and submesothelial cells (arrow). Some desmin ϩ HSCs and p75Ntr ϩ Sma ϩ perivascular mesenchymal cells (arrows) express DLK1. The nuclei were counterstained with DAPI. (Original blots that generated this figure are shown in suppl. Fig. S1.) C, FACS of GFP ϩ cells from E13.5 embryonic or adult livers of the GFP-Coll mice. WT littermates were used as negative controls. GFP ϩ populations were sorted for mRNA expression analysis. D, quantitative PCR of liver cells separated by FACS. The expression values were normalized against Gapdh. The GFP ϩ populations from E13.5 or adult livers enrich expression of desmin and Col1a1. The E13.5 GFP ϩ cells strongly express Dlk1, Lhx2, and Ptn compared with the adult GFP ϩ cells. **, p Ͻ 0.01 compared with the E13.5 GFP ϩ cells. post-PH or a possible compensatory mechanism which has restored a normal regenerative activity. Nonetheless, our study is the first to identify the pivotal role of HSC-derived DLK1 in promoting HC proliferation and liver regeneration during the early phase following PH most likely via the DLK-Wnt10b pathway.
Our DLK1 expression analysis in embryonic livers also demonstrates the expression of this protein by HSCs in chicken embryos and by HSC precursors (mesothelial and submesothelial cells) and HSCs in mouse embryos along with the expression by hepatoblasts, establishing an observation parallel to what we have shown in adult liver regeneration. This raises an intriguing possibility that paracrine-autocrine loops of DLK1 may also participate in liver development. It is also possible that hepatocyte-derived DLK1 induces DLK1 expression by HSCs in liver regeneration. This paradigm has recently been proposed for CCl 4 -induced liver fibrosis (44). In that study, no or minimal expression of Dlk1 mRNA is detected in HSCs in sharp contrast to higher levels of expression in HSCs than hepatocytes isolated at day 1 after PH (Fig. 4B). HSCs isolated from CCl 4 -induced and cholestatic liver fibrosis show induction of Dlk1 mRNA (Fig. 1C). Our cross-culture experiment also demonstrates induction of Dlk1 in hepatocytes from Sham-operated mice (which express no or minimal Dlk1) co-cultured with HSCs from PH mice expressing abundant Dlk1 and abrogation of this effect with anti-DLK1 antibody (Fig. 5D). Thus, we present compelling evidence that HSCs up-regulates Dlk1 in activation in culture as well as fibrosis and regeneration models. Reasons for this discrepancy are not known presently.
Global Dlk1 knock-out mice may be considered for testing this morphogen in liver development and regeneration. Dlk1 knock-out mice have pre-and postnatal growth retardation with skeletal malformation (45), but liver phenotype has not been examined. These mice also exhibit hyperlipidemia and increased adiposity (45). Because these metabolic complications may present confounding effects on assessment of liver regeneration, cell type-specific Dlk1 ablation appears a most logical approach that would also provide more insights into HSC-hepatocyte cross-talk.
In conclusion, our study demonstrates for the first time that HSCs are the primary source of DLK1 in normal adult liver and Dlk1 induction has a causal role in myofibroblastic trans-differentiation of HSCs via epigenetic repression of Ppar␥ involving canonical Wnt pathway. DLK1 induced in activated HSCs has a growth promoting effect on HC in liver regeneration via its paracrine and self-inductive interactions with the target cells. Detailed investigation of DLK1-mediated mesenchymal-epithelial interactions may disclose potential therapeutic targets to maximize beneficial effects on chronic liver disease.