DNA methylation maintenance at the p53 locus initiates biliary-mediated liver regeneration

In cases of extensive liver injury, biliary epithelial cells (BECs) dedifferentiate into bipotential progenitor cells (BPPCs), then redifferentiate into hepatocytes and BECs to accomplish liver regeneration. Whether epigenetic regulations, particularly DNA methylation maintenance enzymes, play a role in this biliary-mediated liver regeneration remains unknown. Here we show that in response to extensive hepatocyte damages, expression of dnmt1 is upregulated in BECs to methylate DNA at the p53 locus, which represses p53 transcription, and in turn, derepresses mTORC1 signaling to activate BEC dedifferentiation. After BEC dedifferentiation and BPPC formation, DNA methylation at the p53 locus maintains in BPPCs to continue blocking p53 transcription, which derepresses Bmp signaling to induce BPPC redifferentiation. Thus, this study reveals promotive roles and mechanisms of DNA methylation at the p53 locus in both dedifferentiation and redifferentiation stages of biliary-mediated liver regeneration, implicating DNA methylation and p53 as potential targets to stimulate regeneration after extensive liver injury.


INTRODUCTION
The liver is the largest digestive organ and performs multiple physiological functions, such as metabolism and detoxification, with remarkable regenerative capacity. The hepatic parenchyma cells, including hepatocytes and cholangiocytes, enlarge and proliferate for restoring hepatic mass upon mild liver injury after partial hepatectomy (PH) and most of the drug-induced liver injury 1 . In case of severe liver injury, such as end-stage liver diseases, when the proliferation of existing hepatocytes is compromised, the only treatment is liver transplantation, but the donors are limited. Using zebrafish as liver injury model, we and others found that the biliary epithelial cells (BECs) dedifferentiate into bipotential progenitor cells (BPPCs) expressing the endoderm and hepatoblast markers and then redifferentiate into hepatocytes and BECs after extensive hepatocyte damages [2][3][4] . This injury-induced BECs-hepatocytes trans-differentiation in zebrafish is conserved in mammalian mice when blocking hepatocyte proliferation or long-term chronic liver injury [5][6][7][8][9] . Promoting the trans-differentiation from BECs into hepatocytes may have potential clinical significance for patients with severe liver diseases.
DNA methylation is an essential epigenetic mechanism regulating gene expression and organ regeneration 26 . Three enzymatically DNA methyltransferases, including DNMT1, DNMT3A, DNMT3B, have been found in the vertebrate. Whereas DNMT3A and DNMT3B are essential for de novo DNA methylation 27 , DNMT1 is the primary enzyme responsible for maintaining DNA methylation during cell division 28 . Importantly, 5-azacytidine (5azaC), a specific inhibitor of DNMT1, has been widely used to inhibit DNA methylation 29 . Several studies show that DNA methylation is involved in organ regeneration, including pancreas regeneration 30 , axonal regeneration 31,32 , and retina regeneration 33 . Liver-specific Dnmt1 knockout in mice demonstrated that DNMT1 plays an indispensable role in the genomic stability of hepatocytes in growth and the cell survival of liver progenitors 34 . Besides, liver-specific Uhrf1 deletion induces DNA hypomethylation, activates the pro-regenerative genes, and enhances liver regeneration after PH 35 . Furthermore, DNA demethylation caused by TET1 licenses adult cholangiocytes for organoid formation 24 . Although the roles of DNA demethylation have been reported in different liver injury models, the functions of DNMT1 and DNA methylation maintenance in biliary-mediated liver regeneration have not been investigated.
In this study, we explored the roles of DNA methylation maintenance by Dnmt1 in BECs-mediated liver regeneration. We used DNA methylation inhibitor and dnmt1 mutant to address the BEC dedifferentiation and BPPC redifferentiation. We uncovered that the maintenance of DNA methylation at the p53 locus promotes the BEC dedifferentiation through derepressing mTORC1 signaling and induces BPPC redifferentiation through derepressing Bmp signaling. Loss of dnmt1 blocks liver regeneration. Furthermore, DNA methylation level is maintained in hepatic progenitor cells in mice fed with a CDE diet.

RESULTS
The maintenance of DNA methylation is required for the BECmediated liver regeneration Dnmt1 maintains DNA methylation during cell division and proliferation 28 . However, the roles of DNA methylation in liver regeneration after severe hepatocyte damages are unknown. To explore the roles of DNA methylation in liver regeneration after extensive hepatocyte loss, we applied the Nitroreductase-Metronidazole (NTR-Mtz)-based zebrafish liver injury model 2,3 and detected the expressions of DNA methyltransferases, including dnmt1, dnmt3aa, dnmt3ab, and dnmt3bb.1 (Supplementary Fig. 1). The expression of dnmt1, a DNA methylation maintenance enzyme, was specifically upregulated in the regenerating livers from regeneration 0 h (R0h) to R24h ( Supplementary Fig. 1a). While the DNA de novo methylation enzymes, such as dnmt3aa, dnmt3ab, and dnmt3bb.1, were not expressed during liver regeneration ( Supplementary Fig. 1b-1d). Anxa4, a BEC-specific marker 36 , labels BECs and the BECs-derived bipotential progenitor cells (BPPCs) during liver regeneration 2,3,12 . Whole-mount antibody staining further validated the expression of Dnmt1 specifically in the Anxa4-positive BPPCs from R0h to R24h and weak expression at R48h (Fig. 1a). To study DNA methylation levels during liver regeneration, we analyzed the expression of 5-methyl Cytosine (5meC), an indicator of DNA methylation 37 . In the Tg(tp1:Tomato; lfabp:Dendra2-NTR) transgenic line in which the BECs and BPPCs were labeled by Tomato, the expression of 5meC maintained in the Tomato-positive BPPCs from R0h to R48h (Fig. 1b, arrows). To explore the roles of Dnmt1 and DNA methylation in liver regeneration, 5-azacytidine (5azaC), a specific inhibitor of Dnmt1 29 , was applied at different stages of liver regeneration ( Fig. 1c and Supplementary Fig. 2a). The 5azaC treatment significantly reduced the expression levels of 5meC in BPPCs ( Supplementary Fig. 2b). The larvae treated with 10 mM Mtz from 5 days post-fertilization (dpf) (before treatment, BT) to 6 dpf for 24 h regenerated normal liver with functional hepatic markers ceruloplasmin (cp) and vitamin D binding protein (gc) 2,3,12 at R48h (Fig. 1c and Supplementary Fig. 2c). Consequently, after 5azaC treatment from 5 dpf/BT to 6 dpf/R0h (early inhibition of DNA methylation), liver regeneration hardly occurred, as exhibited by the Dendra2 epifluoscence or expressions of functional hepatocyte markers cp and gc 2,3,12 ( Fig. 1c and Supplementary Fig. 2c). While 5azaC treatment from R0h to R48h (late inhibition of DNA methylation) led to small regenerating livers ( Fig. 1c and Supplementary Fig. 2c). The morphology of larvae remained normal after 5azaC treatment, excluding any side effects on liver regeneration ( Supplementary Fig. 2d). These data indicate that the maintenance of DNA methylation plays important roles in liver regeneration.
DNA methylation maintenance governs BEC dedifferentiation at the early stage of liver regeneration To investigate the role of DNA methylation maintenance in the initiation stage of liver regeneration, we used the protocol of early 5azaC incubation during Mtz treatment from 5 dpf/BT to 6 dpf/ R0h for 24 h (early inhibition of DNA methylation) ( Fig. 2a and Supplementary Fig. 4a). Given that BECs firstly dedifferentiate into BPPCs after liver damage, with activations of endoderm and hepatoblast developmental markers foxa3, hhex, and sox9b 3,12 , we checked the effects of 5azaC on the BEC dedifferentiation at R0h. Fluorescent in situ hybridization (FISH) coupled with antibody staining 38 showed that the transcriptional activations of foxa3, hhex, and sox9b were reduced at R0h and R8h in the 5azaC-treated groups compared to the controls (Fig. 2b), but became recovered at R16h (Supplementary Fig. 4b). These data indicate the delayed BEC dedifferentiation by the early inhibition of DNA methylation.
The mTORC1 signaling plays critical roles in activating BEC dedifferentiation after extensive hepatocyte damages 12 . Downregulation of DNMT1 has been reported to decrease the activation of mTORC1 signaling in kidney disease 39 . Therefore, we hypothesized that the activity of mTORC1 is regulated by DNA methylation. The phosphorylated ribosomal S6 protein (pS6) and phosphorylated 4E-binding protein 1 (p4EBP1), two wellestablished downstream effectors of mTORC1 signaling 40 , were analyzed. The expressions of pS6 and p4EBP1 were inhibited from R0h to R8h, but became recovered at R16h after early inhibition of DNA methylation ( Fig. 2b and Supplementary Fig. 4b). These results suggest that early inhibition of DNA methylation delays the activation of mTORC1 signaling, indicating important roles of DNA methylation in the initiation of liver regeneration.
DNA methylation maintenance at the p53 locus promotes BEC dedifferentiation through derepressing mTORC1 signaling Dnmt1 has been reported to maintain pancreatic progenitor cell survival and regeneration by negatively regulating p53 30,41 . To investigate whether p53 was aberrantly activated after early inhibition of DNA methylation, we examined DNA methylation at the p53 locus at R0h with the analysis by MethPrimer 42 and QUMA 43 ( Fig. 3a and b). In contrast to BT and R0h groups, 5azaC treatment led to significantly reduced level of DNA methylation at the p53 locus (Fig. 3b). Thus, transcription of p53 was upregulated from R0h to R48h ( Fig. 3c; arrows). These results indicate that early inhibition of DNA methylation activates p53 transcription during liver regeneration.
Aberrant activation of p53 reduces the mTOR signaling in the setting of genotoxicities and stresses [44][45][46] . To explore whether DNA methylation regulates BEC dedifferentiation through p53, we analyzed hepatoblast markers and mTORC1 effectors in the p53 mutant (Fig. 3d). The expressions of hhex, foxa3, sox9b, pS6, and p4EBP1 at R8h after early inhibition of DNA methylation were partially rescued in the p53 mutant compared to the wild-type (WT) (Fig. 3d). The DNA methylation levels at the loci of foxa3, hhex, and sox9b were unaffected after 5azaC treatment, while the DNA methylation at the p53 locus remained downregulated in the p53 mutant after 5azaC treatment ( Supplementary Fig. 5), indicating that DNA methylation at the p53 locus is specifically regulated in the context of liver regeneration. These data indicate that DNA methylation at the p53 locus derepresses the mTORC1 signaling and induces BEC dedifferentiation.
To explore the effects of early inhibition of DNA methylation at the p53 locus on liver regeneration, we checked the expressions of hepatocyte and BEC markers at R48h in the p53 mutant. Mutation of p53 rescued the loss of cp and gc expressions caused by 5azaC treatment ( Fig. 4a and b). The expression of Bhmt and the ratio of Alcam + Dendra2 − mature BECs also recovered in the p53 mutant (Fig. 4c). As a consequence, the reduced size of regenerating liver caused by the early incubation of 5azaC was partially rescued in the p53 mutant ( Supplementary Fig. 6a-c). The rescued hepatocyte regeneration by p53 mutation after early 5azaC treatment did double transgenic line from BT to R48h. Note that the expression of 5meC maintains in Tomato+ cells from R0h to R24h (arrows). c Confocal projection images showing the liver regeneration from BT to R48h after 5azaC treatment from BT to R0h (early DNA methylation inhibition) and R0h to R48h (late DNA methylation inhibition). Quantification of the area of liver sizes and the intensity of Dendra2 expression at R48h. Asterisks indicate statistical significance: *P < 0.05; ****P < 0.0001 using t-tests analysis when compared to control. Numbers indicate the proportion of larvae exhibiting the expression shown. Scale bars: 100 µm; error bars, ±SEM. DAPI 4' , 6-diamidino-2-phenylindole, BT before treatment, R regeneration time after the withdrawal of Mtz.
not affect the origin of nascent hepatocytes, which were also derived from BEC trans-differentiation ( Supplementary Fig. 7). All these data indicate that DNA methylation at the p53 locus regulates BEC-mediated liver regeneration.
Maintenance of DNA methylation at the p53 locus promotes BPPC redifferentiation at late stages of liver regeneration To evaluate whether DNA methylation is also involved in the late stages of liver regeneration, we used the protocol of late 5azaC treatment from R0h to R48h (late inhibition of DNA methylation) to bypass its earlier effects on BEC dedifferentiation (Fig. 5a). Late inhibition of DNA methylation reduced the sizes of regenerating liver at R24h and R48h (Supplementary Fig. 2c and 6b). Then, we analyzed whether these late effects on liver regeneration also involved by p53. Late 5azaC treatment reduced the level of DNA methylation on the p53 locus at R8h (Fig. 5b), leading to elevated p53 expression in the regenerating liver from R8h to R48h ( Fig. 5c; arrows). To assess whether DNA methylation affects the BPPC redifferentiation through p53, we checked liver regeneration at R48h in the p53 mutant ( Fig. 5d and Supplementary Fig. 6a). The sizes of regenerating livers, the expression of hepatocyte makers cp, gc, and Bhmt, and the number of Alcam + Dendra2mature BECs reduced at R48h after late 5azaC treatment. By contrast, in the p53 mutant, the sizes of regenerating livers and the redifferentiation from BPPCs to hepatocytes and BECs were partially rescued ( Fig. 5e and f and Supplementary Fig. 6b and  6c). Similarly, all the rescued regenerating hepatocytes in the p53 mutant were also derived from BEC trans-differentiation after late 5azaC treatment ( Supplementary Fig. 7). These data indicate that loss of p53 could rescue the defective liver regeneration caused by late inhibition of DNA methylation.  Early DNA methylation inhibition reduces BEC dedifferentiation through p53 activation. a Experimental scheme illustrating the 5azaC and Mtz treatment and analysis from R0h to R48h. b Bisulfite sequencing analysis of DNA methylation at CpG islands near the promoter at the p53 locus in DNA isolated from DMSO and 5azaC-treated livers at BT and R0h. c WISH images showing the expression of p53 during liver regeneration after early DNA methylation inhibition. Quantification of the intensity of p53 expression in liver regions from R0h to R48h. d Single-optical section images showing the expressions of hepatoblast markers hhex, foxa3, sox9b, and mTORC1 effectors pS6 and p4EBP1 at R8h in WT and p53 mutant after early DNA methylation inhibition. Note that the expressions of hepatoblast markers and mTORC1 signaling were rescued in p53 mutant compared with WT after early DNA methylation inhibition. Asterisks indicate statistical significance: ****P < 0.0001 using t-tests analysis when compared to control. Numbers indicate the proportion of larvae exhibiting the expression shown. Scale bars: 100 µm; error bars, ±SEM. DAPI 4′, 6-diamidino-2-phenylindole, BT before treatment, FISH fluorescent in situ hybridization, R regeneration time after the withdrawal of Mtz.
DNA methylation at the p53 locus activates BPPC redifferentiation through derepressing Bmp signaling Bmp signaling governs BPPC redifferentiation through tbx2b, id2a, and smad5 after extensive liver damages in zebrafish 20 . p53 regulates neural stem cell differentiation through BMP signaling 47 . So, we tested whether loss of DNA methylation at the p53 locus could affect Bmp signaling during BPPC redifferentiation. The transcriptional activations of id2a, tbx2b, and smad5, three critical Bmp target genes 20 , were repressed in the regenerating livers at R24h after late 5azaC treatment, which were partially rescued in the p53 mutant (Fig. 6a). To validate whether the effects of late 5azaC treatment on BPPC redifferentiation are mediated by the Bmp signaling, 5azaC was applied from R0h and bmp2b overexpression was induced by heat shock from R8h under the Tg (hsp70l:Bmp2b) 48 transgenic background (Fig. 6b). Overexpression of bmp2b rescued the reduced size of regenerating liver in the 5azaC-treated group, but hardly accelerated normal liver regeneration in the control group (Fig. 6c). The expressions of mature hepatocyte and BEC markers were all rescued in the bmp2b overexpression groups after 5azaC treatment ( Fig. 6d and e). In addition, the expressions of foxa3, hhex, and sox9b, as well as the mTORC1 effectors pS6 and p4EBP1, remained relatively normal at R24h after late 5azaC treatment, indicating that late inhibition of DNA methylation is ineffective to mTORC1 activation and BEC dedifferentiation ( Supplementary Fig. 8). All these results indicated that DNA methylation at the p53 locus promotes BPPC redifferentiation through Bmp signaling.
The dnmt1 mutant exhibits aberrant p53 activation and impaired BEC dedifferentiation and BPPC redifferentiation To further substantiate the roles of DNA methylation in liver regeneration, we checked the liver regeneration in the dnmt1 s872 mutant 30 . The dnmt1 mutant displays liver degeneration 30 from Fig. 4 Loss of p53 rescues liver regeneration defect after early DNA methylation inhibition. a Experimental scheme illustrating the 5azaC and Mtz treatment and analysis at R48h. b WISH images showing the expressions of hepatocyte markers cp and gc at R48h in WT and p53 mutant after early DNA methylation inhibition. Note that the expressions of cp and gc were recovered in p53 mutant after 5azaC treatment. Quantification of the intensity of cp and gc expression in liver regions at R48h. c Single-optical section images showing the expressions of hepatocyte marker Bhmt and BECs marker Alcam in WT and p53 mutant after early DNA methylation inhibition and controls at R48h. Quantification of the intensity of Bhmt expression and the percentage of Alcam+ among Dendra2+ cells in liver regions at R48h. Note that the expression of Bhmt was recovered in p53 mutant compared to WT after 5azaC treatment. Asterisks indicate statistical significance: ***P < 0.001; ****P < 0.0001 using t-tests analysis when compared to control. Scale bars: 100 µm; error bars, ±SEM. DAPI 4′, 6-diamidino-2phenylindole, BT before treatment, ns no significant difference, R regeneration time after the withdrawal of Mtz. Note that the downregulation of cp and gc expressions was recovered in p53 mutant after 5azaC treatment. Asterisks indicate statistical significance: ***P < 0.001; ****P < 0.0001 using t-tests analysis when compared to control. Scale bars: 100 µm; error bars, ±SEM. DAPI 4′,6-diamidino-2phenylindole, BT before treatment, ns no significant difference, R regeneration time after the withdrawal of Mtz. Fig. 9a). To exclude the side effect of liver degeneration, we ablated the hepatocytes with Mtz from 4 dpf for 24 h, then checked the liver regeneration. Most of the dnmt1 mutants survived to R48h, but liver regeneration was defective ( Supplementary Fig. 9a and 9b). The expression level of 5meC in Tomato-positive BPPCs was downregulated in the dnmt1 mutant at R0h (Supplementary Fig. 9c). To exclude the abnormal liver development at late stages of liver regeneration, we injected dnmt1 mRNA into the dnmt1 mutant, then examined liver development and regeneration from 4 dpf to R48h (Supplementary Fig. 10a). Because the injected mRNA could maintain in the embryo only for 3-4 days, the injected dnmt1 mRNA partially rescued the defective liver development at 4 dpf, but failed to rescue later liver regeneration (Supplementary Fig. 10). The dedifferentiation was arrested in the dnmt1 mutant as assessed by the expression of mTORC1 effectors and hepatoblast markers at R0h (Fig. 7a and b). The DNA methylation level at the p53 locus was reduced, and the transcriptional level of p53 was upregulated in the dnmt1 mutant at R0h (Fig. 7c and d). These data indicate that dnmt1 is important for BEC dedifferentation.

dpf (Supplementary
To further access the roles of dnmt1 in the late stages of liver regeneration, we analyzed BPPC redifferentiation in the dnmt1 mutant. The expressions of hepatocyte markers including cp, gc, and Bhmt, as well as the BEC marker Alcam, were undetectable in the dnmt1 mutant at R48h (Fig. 7e and f). The expressions of p53 itself, as well as its target genes p21 and mdm2, were upregulated Fig. 6 Late DNA methylation inhibition activates p53 and in turn reduces BPPC redifferentiation through Bmp signaling. a WISH images showing the expressions of id2a, tbx2b, and smad5 at R24h treated with DMSO or 5azaC in p53 mutant and wild-type livers (arrows). Quantification of the intensity of id2a, tbx2b, and smad5 expressions in liver regions at R24h. Note that the expressions of id2a, tbx2b, and smad5 were recovered in p53 mutant compared with control after 5azaC treatment. b Experimental scheme illustrating the 5azaC and Mtz treatment in transgenic line Tg(lfabp:Dendra2-NTR; hsp70l:Bmp2b) and heat shock from R8h to R48h. c Confocal images showing the regenerating livers at R48h after heat-shock and 5azaC treatment. d WISH images showing the expressions of gc and cp at R48h after heatshock and 5azaC treatment. Quantification of the intensity of gc and cp expressions in liver regions at R48h. e Single-optical section images showing the expressions of hepatocyte marker Bhmt and BECs marker Alcam at R48h after heat-shock and 5azaC treatment. Quantification of the intensity of Bhmt expression and the percentage of Alcam+ among Dendra2+ cells in liver regions at R48h. Asterisks indicate statistical significance: **P < 0.01; ***P < 0.001; ****P < 0.0001 using t-tests analysis when compared to control. Numbers indicate the proportion of larvae exhibiting the expression shown. Scale bars: 100 µm; error bars, ±SEM. DAPI 4′,6-diamidino-2-phenylindole, BT before treatment, HS heat shock, OE overexpression, R regeneration time after the withdrawal of Mtz. in the liver of dnmt1 mutant from R0h to R48h (Fig. 7g and Supplementary Fig. 11). Taken together, these results suggest that dnmt1 is important for both BEC dedifferentiation and BPPC redifferentiation through the inactivation of p53.

Aberrant activation of p53 impairs the BEC-mediated liver regeneration
To determine the roles of p53 in liver regeneration, we used the Tet-on system 49 to specifically overexpress p53 in BECs under the Tg(Krt18:Tet3G; Tre3G:p53-2A-Venus; lfabp:Dendra2-NTR; Tp1:GFP) transgenic background (Fig. 8a). The Doxycycline (Dox) treatment from 4 dpf to BT induced overexpression of p53 in the Tp1:GFP + BECs at BT, which led to reduced liver regeneration at R48h (Fig. 8b-e). General overexpression of p53 was achieved by heat shock under the Tg(hsp70l:p53-2A-Venus) transgenic background ( Supplementary Fig. 12a and 12b), which led to defective liver regeneration ( Supplementary Fig. 12c). The p53 overexpression from BT to R0h reduced the mTORC1 signaling and the expression of hepatoblast markers at R0h (Supplementary Fig. 13a-c), indicating impaired BEC dedifferentiation and BPPC formation. Furthermore, overexpression of p53 from R0h to R24h repressed Bmp signaling at R24h (Supplementary Fig. 13d and 13e). These data demonstrate that aberrant p53 activation impairs early and late stages of liver regeneration through mTORC1 and Bmp signaling, respectively.
The maintenance of DNA methylation is essential for ductular reaction in mice The roles of DNA methylation in BEC dedifferentiation in zebrafish prompted us to check its effects in mice. We made a pilot study using a choline-deficient, ethionine-supplemented (CDE) diet to investigate the ductular reaction in mice, in which hepatic progenitor cells are formed from BECs 50 . Mice fed with CDE diet for 10 days were subjected to 5azaC or PBS injection from day 3 to day 9 ( Supplementary Fig. 14a). Liver damage was comparable between vehicle and 5azaC-injected mice as indicated by Serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), and bilirubin ( Supplementary Fig. 14b). The level of DNA methylation was high in BECs in the chow control, and maintained in the KRT19 + ductular cells after liver injury (Supplementary Fig.  14c). The 5azaC treatment significantly reduced the level of DNA methylation in the ductular cells, but not in other hepatic cells ( Supplementary Fig. 14c). To verify the effects of DNA methylation on the formation of progenitors, we examined the expression patterns of SOX9, a well-known hepatic progenitor cell marker 50 and KRT19. The expression areas of SOX9 and KRT19 were obviously reduced in the 5azaC-injected mice ( Supplementary Fig.  14d), indicating the importance of DNA methylation in ductular reaction. Altogether, these mouse data indicate that the maintenance of DNA methylation regulates the formation of progenitor cells as in zebrafish.

DISCUSSION
Zebrafish as an animal model is widely used to explore the cellular and molecular mechanisms underlying organ development and tissue regeneration 3,51-57 . For the zebrafish, extensive hepatocyte damages and mouse extensive liver injury models allow the evaluation of cellular and molecular mechanisms during liver regeneration in end-stage liver diseases. Uncovering the detailed mechanisms underlying the BEC dedifferentiation and redifferentiation are crucial to promoting BECs-mediated liver regeneration. Here, we describe that DNA methylation inhibition from BT-R0h and dnmt1 mutants reduces the dedifferentiation of BECs with the activation of p53 after liver injury. Moreover, DNA methylation inhibition from R0h to R48h reduces the BPPC redifferentiation. Hypermethylation of p53 regulates the dedifferentiation of BECs by mTORC1 signaling and induces BPPC redifferentiation through Bmp signaling. Loss of p53 rescues BPPCs activation and redifferentiation in liver regeneration after inhibition of DNA methylation (Fig. 8f). Furthermore, we show that DNA methylation may play a role in mouse hepatic progenitor cell activation.
Epigenetic regulation has an essential role in the progress of liver regeneration. Histone deacetylation enzyme (Hdac1) regulates differentiation of BPPCs after extensive hepatocyte damages in zebrafish 23 . The DNA methylation factor DNMT1 maintains the liver homeostasis and regeneration after partial hepatectomy (PH) in mice 34 . Another DNA methylation regulator, Uhrf1, mediates liver regeneration after PH in zebrafish 58 and mice 35 . Although DNA methylation is vital for liver development 30,58,59 , we found the dedifferentiation of BECs, the redifferentiation of BPPCs were blocked in dnmt1 mutant and DNA methylation inhibition. Indeed, knockdown of Dnmt1 in cultured hepatic progenitor cells leads to sphere formation defect 34 . Although the other hepatic cells may also express high level of DNA methylation and play roles in liver regeneration, we found Dnmt1 is specifically upregulated in BPPCs after liver injury. These data indicate Dnmt1 regulates liver regeneration in a cellautonomous manner. The DNA methylation levels are slightly downregulated from R0h to R24h in BPPCs, which validates that the DNA hypomethylation of some other genes may also play essential roles during liver regeneration 24 , suggestings that different genes with distinct DNA methylation levels on their promoters contribute to liver regeneration together.
The dedifferentiation of BECs and the activation of mTORC1 signaling are reduced after early DNA methylation inhibition, loss of p53 partially rescues the expressions of mTORC1 signaling and hepatoblast markers in BPPCs at R8h. Based on the crucial roles of mTORC1 signaling in the early stage of liver regeneration 12,14 , other compensation pathways may involve in regulating the mTORC1 signaling after DNA hypomethylation. Indeed, epigenetic compensation has been reported to promote liver regeneration after DNA methylation inhibition 35 . Some stem genes and regeneration pathways may be activated to induce liver regeneration after DNA hypomethylation 24 . The factors, such as sesn1/2, cab39l, and ddit3/4, may play roles in inhibiting the activation of mTORC1 signaling 39,46,60 . P53 and Ddit4 synergetically license differentiated cells to the cell cycle by suppressing mTORC1 signaling in tissue regeneration 60 . It may be the reason why the only loss of one gene, such as p53, partially rescues the depression of mTORC1 signaling after inhibition of DNA methylation. Accordingly, p53 activation negatively regulates zebrafish heart regeneration 61 . Our results indicate the mechanisms by which Dnmt1 mediates BPPCs activation, in part, through the regulation of p53 signaling. Whether other factors that Dnmt1 transcriptionally represses are involved in this progress needs further exploration.
Dnmt1 controls the expression of p53 both in organ development and regeneration 30,41 . Inhibition of DNA methylation from BT to R0h and R0h to R48h activates the expression of p53. The roles of p53 in cell apoptosis have been well-known 34,41 , but whether p53 affects biliary-derived liver regeneration is less addressed. The liver stem cells develop into hepatocellular carcinoma in p53-deficient mice 62 , and our results show normal liver regeneration in zebrafish p53 mutant. This discrepancy may be caused by the short time of liver regeneration (about 48 h) in zebrafish liver regeneration. BEC-specific and whole-body overexpression of p53 reduce liver regeneration, indicating p53 regulates biliary-derived liver regeneration in a cell-autonomous manner. The sizes of regenerating liver and the expressions of hepatocyte markers decreased after strong genetic overexpression of p53. Indeed, p53 overexpression during liver injury reduces the activity of mTORC1 in BPPCs, and p53 overexpression in BPPC redifferentiation reduces Bmp signaling, suggesting p53 regulation is stage-dependent. p53 has been reported to regulate BMP signaling 47,63 . However, how p53 represses mTORC1 and Bmp signaling during liver regeneration needs to be further explored.
Above all, this study explored the roles of DNA methylation in BECs-mediated liver regeneration. DNA methylation at the p53 locus represses p53 transcription and in turn derepresses mTORC1 signaling to activate BEC dedifferentiation. After BEC dedifferentiation and BPPCs formation, DNA methylation at the p53 locus maintains in BPPCs to continue blocking p53 transcription, which derepresses Bmp signaling to induce BPPC redifferentiation. This study indicates the roles of DNA methylation and p53 as potential targets for therapeutics after severe acute liver injuries.

METHODS
The agency or committee that granted approval All animals are housed in a temperature-and light-controlled facilities and are maintained in accordance with the Guide for Care and Use of Laboratory Animals and the Animal Welfare Act. Experiments were performed with the approval of the Institutional Animal Care and Use Committee (IACUC) at Southwest University.

Dnmt1 mRNA microinjection
Zebrafish dnmt1 full-length coding sequence was cloned from 3 dpf cDNA. The dnmt1 PCR product was subcloned into the pCS2(+) vector by the NovoRec PCR step directional cloning kit (Novoprotein, China) for RNA preparation. Sense mRNA was synthesized from a linearized plasmid template using the mMESSAGE mMACHINE Kit (Ambion). 150 pg dnmt1 mRNA injected into the cell of the one-cell stage embryos. The primers used in this study are listed in Supplementary Table 2.

Mouse studies
Four-week-old female C57BL/6-WT mice were housed under pathogen-free conditions under 12 h light-dark cycles. All experiments were performed following the guidelines of the IACUC at Southwest University. The mice were given Choline Deficient Chow (XieTong Biotech, Nanjing, China) and drinking water containing 0.15% (wt/vol) DL-ethionine (TCI). After 3 days on the CDE diet, animals received daily intraperitoneal injections of the 5azaC (Selleck) dissolved in the PBS at a dose of 2.5 µg/g. After 7 days of injections, the mice were sacrificed and liver tissue was collected and fixed in 4% PFA for analysis.
For immunohistochemical staining, the mouse liver tissue was embedded in 4% low melting point agarose; then the tissue was sliced into 40 µm sections. The sections underwent antigen retrieval in sodium citrate buffer (pH 6.0) or Tris-EDTA buffer (pH 8.5) in a Dry Bath Incubator for 10 minutes before blocking and permeabilization. The sections were permeabilized using 1% Triton X-100/PBS for 2 h at room temperature, followed by blocking with PBTN for 1 h at room temperature. Then the sections were incubated with the KRT19 primary antibody (1:500; ab133496; Abcam, Cambridge, MA), SOX9 primary antibody (1:300; A5080, Bimake), and 5meC primary antibody (1:100; ab10805, Abcam, Cambridge, MA) overnight at 4°C. (For immunohistochemical staining of 5meC primary antibody, sections were incubated with 2 N HCl overnight at room temperature and then neutralized in 100 mM Tris-HCl (pH 8.5) for 1 h).

Whole-mount in situ hybridization
For Whole-mount in situ hybridization, cDNA from the whole embryos at 3 dpf, R0h, and R48h were used as a template to amplify the genes by PCR. Digoxygenin (DIG) labeled antisense probes were synthesized from PCR products using the T7 or SP6 RNA polymerase (Roche). Larvae at the desired stages were fixed with 4% paraformaldehyde (PFA) in PBS at 4°C overnight, then dehydrated by methanol and stored at −20°C overnight. The larvae were then rehydrated successively in 75, 50, 25, and 0% methanol in PBT (0.1% Tween-20 in PBS), then prehybridized the larvae in HYB (50% formamide, 5 × SSC, 0.1% Tween-20, 5 mg/ml torula yeast RNA, 50 mg/ml heparin) at 68.5°C for 3-5 h. The HYB was replaced with antisense probes diluted in HYB and hybridized at 68.5°C overnight. The larvae were gradually washed at 68.5°C with 100, 75, 50, and 25% HYB in 2 × SSCT, and finally in 0.2 × SSCT. Then the larvae were progressively replaced with 25, 50, 75, and 100% MABT (150 mM maleic acid, 100 mM NaCl, 0.1% Tween-20, pH 7.5) in 0.2 × SSCT, and blocked in 2% Block Reagent (Roche) at room temperature for 2 h. Larvae were then incubated with anti-Dig-AP antibody (Roche) diluted at 1/2000 with blocking buffer at 4°C overnight. Larvae were rinsed with MABT for eight times, then staining with NBT/BCIP solution (Roche). The WISH images were captured using a SteREO DiscoveryV20 microscope equipped with AxioVision Rel 4.8.2 software (Carl Zeiss, Jena, Germany). The primers are listed in Supplementary Table 2.

Fluorescence in situ hybridization coupled with antibody staining assays
For fluorescence in situ hybridization couple antibody staining, larvae were fixed with 4% PFA in PBS at 4°C overnight, then incubated in 100% methanol at −20°C for 24 h. Larvae were rehydrated into PBT and the larvae skins were manually peeled off with tweezers. Next, the larvae were washed five times with PBST (1% Triton X-100 in PBS) for 1 h, then prehybridized the larvae in HYB at 65°C for 3-5 h. The HYB was replaced with new HYB containing Digoxigenin-labeled antisense probes and hybridized at 65°C overnight. The following day, the larvae were washed into 2 × SSCT and 0.2 × SSCT at 65°C. Then the larvae were replaced with MABT, and blocked for 2-3 h at room temperature with 2% Block Reagent (Roche) in MABT. Larvae were incubated in Anti-Dig POD antibody solution (Roche) diluted at 1/1000 with blocking buffer at 4°C overnight. The next day larvae were progressively washed with MABT for eight times, with PBST for four times, with PBS twice, and then incubated with TSA Plus Fluorescein Solution (Perkin Elmer) overnight at room temperature with gentle shake. The following day, the larvae were washed with PBST for five times and then performed the antibody staining assays. The primers are listed in Supplementary Table 2.

Quantitative-PCR
Total RNA was extracted from 100 to 200 dissected livers using the TRIzol reagent (Roche). cDNAs were synthesized from total RNA by using the Reverse Transcription Kit (Promega) following the manufacturer's instructions. Quantitative real-time PCRs (qPCRs) were performed using the Fast Start Universal SYBR Green Master (Roche), and normalized by eef1a1 in triplicate. The primers used in this study are listed in Supplementary Table 2.
Genomic DNA isolation, bisulfite conversion, and sequencing Liver genomic DNA was extracted from 100 to 150 dissected livers using lysis buffer containing 100 µg/ml Proteinase K at 55°C for 4 h. Then, genomic DNA was purified with phenol: chloroform: isoamyl alcohol (25:24:1), and ethanol. The purified DNA was performed bisulfite conversion following the manufacturer's instructions with the EpiTect Bisulfite Kit (Qiagen). The promoter sequence was amplified from Bisulfiteconverted DNA using primers designed by MethPrimer. PCR products were purified and cloned into the pGEM-T easy vector (Promega). Eight colonies were sequenced for each sample using M13F sequencing primer. QUMA was used to analyze the sequencing results. The primers used in this study are listed in Supplementary Table 2 Image acquisition and processing The zebrafish larvae were imaged using a SteREO DiscoveryV20 microscope equipped with AxioVision Rel 4.8.2 software (Carl Zeiss). Antibodystained and live larvae were imaged using ZEN 2010 software equipped on an LSM880 confocal microscope (Carl Zeiss). All figures, labels, arrows, scale bars, and outlines were drawn using Adobe Photoshop software.

Quantification and statistical analysis
Automated quantification of the intensity of whole-mount in situ hybridization was performed using ImageJ 1.47 software (National Institute of Health). The fluorescence intensity and the area of liver sizes were quantified with ZEN 2010 (Zeiss). Unpaired two-tailed Student's t-test was used for statistical analysis by GraphPad Prism 7; P < 0.05 was considered statistically significant. Quantitative data were shown as means ± SEM.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.