Hepatocyte regeneration is driven by embryo-like DNA methylation reprogramming

Significance

This paper sheds light on liver regeneration following partial hepatectomy by showing that this process requires programmed changes in DNA methylation that bring about partial dedifferentiation to proliferating progenitor cells no longer present in the adult liver. The identification of this previously unknown pathway suggests that even tissues lacking adult stem cells may be amenable to embryonic reprogramming to create a progenitor state capable of regenerating normal tissue following injury or diseases, such as neurodegeneration or diabetes.

Abstract

As a result of partial hepatectomy, the remaining liver tissue undergoes a process of renewed proliferation that leads to rapid regeneration of the liver. By following the early stages of this process, we observed dramatic programmed changes in the DNA methylation profile, characterized by both de novo and demethylation events, with a subsequent return to the original adult pattern as the liver matures. Strikingly, these transient alterations partially mimic the DNA methylation state of embryonic hepatoblasts (E16.5), indicating that hepatocytes actually undergo epigenetic dedifferentiation. Furthermore, Tet2/Tet3-deletion experiments demonstrated that these changes in methylation are necessary for carrying out basic embryonic functions, such as proliferation, a key step in liver regeneration. This implies that unlike tissue-specific regulatory regions that remain demethylated in the adult, early embryonic genes are programmed to first undergo demethylation, followed by remethylation as development proceeds. The identification of this built-in system may open targeting opportunities for regenerative medicine.

Development involves a progression from early cell types that are multipotent to more definitive tissues that have undergone terminal differentiation. In general, this process is accompanied by irreversible changes in DNA methylation that serve to define and stabilize cell type identity (1). Indeed, according to the Waddington hypothesis (2), each stage of differentiation brings about a loss of developmental potential, which then prevents these specialized cells from dedifferentiation to earlier stages of embryogenesis and it is thought that this may be mediated by DNA methylation. The existence of this type of barrier is what most likely accounts for the inability of most cell types in the body to undergo regeneration when damaged. According to this concept, it would be necessary for the remaining undamaged cells to undergo dedifferentiation back to a stage where they can propagate, thus regenerating enough tissue mass prior to terminal differentiation. Although some tissues in the body constantly maintain a small store of adult stem cells that can easily be recruited to regenerate new cells when needed, in most cases, precursor cells undergo proliferation at an early stage of development but then lose their identity as embryo “stem cells” by undergoing terminal differentiation. The question is whether, for these tissues, we can find a way to induce dedifferentiation to this earlier stage in order to reproliferate progenitor cellular mass, which can then go on to become mature healthy tissue.

One of the most striking examples of cell regeneration takes place in the liver. Following partial hepatectomy (Phx) to remove 2/3 of the liver, the remaining hepatocytes undergo rapid proliferation, thereby regenerating liver mass within 7 d, mainly by hyperplasia (3). Even though the molecular and cellular details of this event have been well documented (4–7), little is known about how this entire process is actually programmed. In this paper, we have investigated the role of DNA methylation in this injury-induced regeneration model. By carrying out genome-wide DNA methylation mapping at progressive stages of regeneration, we have uncovered the existence of an inherent epigenetic program that may play a role in liver development in the embryo but can then be recruited for directing tissue regeneration following injury.

Results

Phx.

In order to determine the epigenetic changes that occur during Phx, we carried out this procedure in 12 wk old mice (female) and then 1 wk later harvested cells from the newly regenerated liver which has already grown back to its original mass. Hepatocytes were purified following liver perfusion, and DNA was prepared for genome-wide bisulfite analysis using the reduced representation bisulfite (RRBS) approach. Although this methodology does not cover all of the CpG sites in the genome, it detects an enriched fraction of regulatory sites (8) and enables one to obtain representative data with excellent read depth. Using a standard pipeline to compile and analyze these results, we could detect 1,684 tiles (100 bp) that were demethylated (>30%) (Fig. 1 A and B), as well as 435 tiles that became de novo methylated (>30%) in the regenerating hepatocytes as compared to normal (Fig. 1 C and D). It is very likely that these sites are representative of nonpromoter regulatory elements, and this was confirmed from Assay for Transposing Accessible Chromatin with sequencing (ATAC-seq) analysis showing that they are specifically activated in posthepatectomy cells (SI Appendix, Fig. S1).

Fig. 1.

DNA methylation changes in hepatocytes following Phx. RRBS data from hepatocytes isolated from 12 wk mice were compared before (Hep) and after partial hepatectomy (HepPhx), and statistically significant DMRs of greater than 30% were identified. These data were then compared to methylation levels (RRBS) for the exact same tiles in implantation embryos (IMP) and hepatocytes isolated 2 mo after Phx (HepPhx*). Tiles undergoing demethylation (A and B) or de novo methylation (C and D) are shown as heatmaps and quantitatively as percent methylation in box plot form, respectively. Boxes include average methylation levels and SD.

In order to put these changes into perspective, we followed the overall developmental history of the differentially methylated regions (DMRs) detected in regenerating hepatocytes. All of these belong to the general class of sequences that originally underwent normal de novo methylation at the time of implantation and then remained in this state in almost all tissues of the body (Fig. 1). The undermethylation observed as a result of Phx thus appears to reflect a specific process of demethylation (Fig. 1 A and B). De novo DMRs, on the other hand, appear to be sequences that specifically underwent demethylation during normal liver differentiation in the late neonatal embryo (Fig. 1 C and D). Thus, the de novo methylation observed following hepatectomy actually constitutes a “reversal” of developmental programming.

These hepatectomy-induced changes in methylation do not appear to be stably maintained with time as shown by analysis of the regenerated liver 9 wk later indicating that all of these methylation patterns have returned to the original state characteristic of normal adult hepatocytes (Fig. 1 B and D). These observations suggest that in contrast to methylation changes that occur as a part of normal liver development, these DMRs represent a transient programmed response to injury which is presumably aimed at allowing the remaining hepatocytes to regenerate. Indeed, almost all of the developmentally generated tissue-specific DMRs, even those that come about postnatally, are highly preserved in these transitory hepatocytes (SI Appendix, Fig. S2).

We next attempted to understand the relationship between the hepatectomy-induced demethylation events and the biology of liver regeneration. Motif analyses of these DMRs [Hypergeometric Optimization of Motif EnRichment (HOMER)] showed a strong enrichment for the transcription factors (TFs), Elk, Etv, and Fli1 (Fig. 2A), all members of the Erythroblast Transformation Specific (ETS) family, which is known to play a role in mediating basic cell functions, such as differentiation, proliferation, survival, as well as several adaptive processes, such as inflammation, wound healing, and tissue remodeling (9–11). A very similar motif pattern was also observed for DMRs resulting from de novo methylation (Fig. 2B). This strong motif profile is very different than that typical of DMRs generated during normal liver maturation, which are highly enriched for hepatocyte-specific TF binding sites (12). In keeping with this, identification of the genes topologically associated with these DMRs by STRING analysis was found to be highly enriched for pathways involved in development and tissue organization as opposed to specific liver function (Fig. 2C). This suggests that the specific epigenetic changes associated with hepatectomy involve very different biological processes.

Fig. 2.

Characterization of regions undergoing Phx-induced methylation changes. The de- (A) and de novo– (B) methylated DMRs shown in Fig. 1 were analyzed for TF motif-binding sites (HOMER), as well as the gene pathways (C) associated with demethylation.

These epigenetic changes were also correlated with gene expression. Analysis of RNA-sequencing (RNA-seq) data (5) identified 148 DMR-associated genes that indeed undergo induction during the first 2 d following Phx (SI Appendix, Fig. S3A). Functional analysis indicated that these DMRs are highly enriched for both ETS and Ap2 TF family motifs (HOMER) and linked to genes (STRING) involved in all aspects of the cell cycle (SI Appendix, Figs. S3 B and C and S4). Taken together, these data suggest that the changes in DNA methylation which occur following Phx may facilitate the alterations needed to enable cell proliferation and growth. Notably, this appears to occur without affecting the terminal hepatocyte-specific methylation pattern (SI Appendix, Fig. S2).

Role of Demethylation in Liver Regeneration.

We next asked whether the changes in DNA methylation that occur as a result of Phx actually play a causative physiological role in the process of regeneration. It has previously been demonstrated that demethylation during development, in particular in the liver, is mediated by the enzymes Tet2 and Tet3 that catalyze the biochemical conversion of 5mC to 5-HmC, which is then converted to cytosine (12–14). Furthermore, deletion of these genes prevents developmentally programmed demethylation, even though this has no effect on the transcription machinery that mediates this process (15). To test the role of demethylation following Phx, Tet2^Flox/Tet3^Flox female mice were injected with adenovirus AAV8 carrying the gene for Cre recombinase under the control of the hepatocyte-specific thyroid-binding globulin (Tbg) promoter 2 wk prior to the Phx procedure, at an age where hepatocytes no longer undergo any further liver-specific demethylation (12).

We isolated hepatocytes 1 wk post-op and first confirmed by PCR analysis that the deletion of all Tet2 and Tet3 alleles was over 90% efficient (12). Furthermore, RRBS analysis of DNA from Tet2/Tet3^−/− hepatocytes demonstrated that this deletion completely prevented the programmed demethylation that usually accompanies Phx (Fig. 3). As a result, the process of liver regeneration was dramatically affected. This can be seen by following the increase in liver mass that occurs during the first few days of regeneration (Fig. 4A). In untreated animals, the liver enlarges by almost threefold during the first 48 h after Phx, with at least part of this increased mass being attributed to hypertrophia (Fig. 4B). By 7 d postsurgery, liver mass reaches its normal level, even though hepatocyte cell size decreases, strongly suggesting that this is accomplished mainly by proliferation. In the Tet-deletion animals, overall regeneration is initially inhibited, but by 7 d posthepatectomy, the liver returns to its normal size (0.04), evidently as a result of hypertrophia (Fig. 4 A–C). It thus appears that the programmed demethylation associated with Phx is probably required mainly but not exclusively for enabling cell proliferation (Fig. 4 D and E). Taken together, these findings strongly indicate that demethylation of specific regulatory regions is required for proper liver regeneration following Phx.

Fig. 3.

Tet2/3 are required for DNA demethylation following Phx. Methylation heatmap (A) and corresponding numerical box plot (B) of Phx–induced demethylation DMRs in wt hepatocytes (average of data from Fig. 1) and Tet2/3-deletion hepatocytes 7 d postinjury.

Fig. 4.

Effect of Tet2/3 deletion on Phx phenotype. (A) Liver to body ratio measured in WT mice and Tet2/3 deletion 2 d after Phx (P-value determined by Welch’s t test). (B) Cell size measurements in WT (blue dots) or Tet2/3-deletion (red dots) hepatocytes before (intact) or after Phx, based on immunofluorescence (IF) anti-β-catenin staining (n = 2 to 6). Significance was determined by the one-way ANOVA test. (C) Examples of IF staining with anti-β-catenin 7 d after Phx in WT as compared to Tet2/3-deletion mice (Scale bar, 100 μm). (D) IHC staining of Ki67 in the intact liver of WT and Tet2/3-deletion mice prior to and 48 h after Phx. (E) Quantitative representation of Ki67 IHC staining (% positive cells) for WT and Tet2/3-deletion intact livers (error bars indicate SEM; n = 3 for each group). P values were calculated using two-way ANOVA.

Liver Regeneration and Development.

In order to better understand the biological processes taking place during hepatocyte regeneration, we used published RNA-seq data (5) taken from cells following Phx and found that this process is accompanied by increased expression of about 1,300 genes as compared to their levels in normal hepatocytes. By means of Genome Ontology analysis, it could be seen that these genes are largely involved in cell-cycle activation and DNA replication, functions that contribute to proliferation (SI Appendix, Fig. S5). Although mature hepatocytes do not divide themselves, they are developmentally preceded in the embryo by a stage of rapid growth. For this reason, we then compared this RNA profile to that of genes up-regulated in hepatoblasts from different stages of liver development (16). Strikingly, they have a very similar pattern of expression as compared to genes that are up-regulated in the embryo with over 70% of the Phx-induced genes also being expressed at elevated levels in E16.5 hepatoblasts (Fig. 5A). A good example of this concept is Sox9, which is known to be expressed in the hepatocyte progenitors (17) and is then induced following Phx (Fig. 5B). These data indicate that the regeneration induced by Phx is initiated by reactivating the same genes needed for early liver development and suggest that the epigenetic process that occurs during Phx may actually mimic a similar developmental program.

Fig. 5.

Regeneration recapitulates early liver development. (A) Venn diagram showing the overlapping expression profile of E16.5 d hepatoblasts as compared to hepatocytes 40 h after Phx (P < 0.0001, one-sample t test). (B) Sox9 staining of normal (Hep) and after Phx. (C) Heatmap showing Phx–induced demethylation (Upper) or de novo methylation (Lower) DMRs, undergoing demethylation or de novo methylation DMRs (Fig. 1) were compared to their methylation levels along development in E11.5 d or E16.5 d hepatoblast DNA, 1 to 3 wk postnatal hepatocytes, as well as an average of Other different adult Tissues (OT, heart, spleen, cerebellum, and muscle).

DNA Methylation Changes during Hepatocyte Development.

Adult hepatocytes are derived from early hepatoblasts, progenitor cells formed from endoderm in 11.5 d embryos. These cells apparently then go through lineage progression over time, ultimately giving rise to both hepatocytes and cholangiocytes. By comparing the DMRs generated by Phx with their methylation state in fetal hepatoblasts, it should be possible to determine whether injury-induced changes may involve epigenetic reprogramming to an earlier stage in liver development.

To this end, we first prepared purified hepatoblasts from 11.5 d embryos, isolated the DNA, and mapped the methylation pattern by RRBS. We then compared this to our Phx-induced DMRs (Fig. 5C). Sites that become de novo methylated following Phx were found to be originally methylated in these early hepatoblasts, so this change indeed appears to constitute a form of epigenetic dedifferentiation. On the other hand, sites that undergo demethylation as a result of Phx are also methylated in hepatoblasts, suggesting that if these sites are indeed subject to demethylation during development, this must occur at a later stage of hepatoblast progression.

In order to test this possibility, we used antibodies that identify specific surface markers to isolate hepatic cells from 16.5 d embryos by Fluorescence Activated Cell Sorting (FACS) and then subjected their DNA to RRBS analysis. We then asked whether the 1,684 tiles that undergo demethylation following Phx are also undermethylated in the embryo. Strikingly, these E16.5 d samples show a large degree of undermethylation specifically at these sites, only to then become remethylated prior to birth (Fig. 5 and SI Appendix, Fig. S3). In contrast, the 435 Phx de novo–methylated sites actually remain methylated during this embryonic stage but then undergo demethylation just prior to birth. These data strongly suggest that the defined epigenetic methylation program activated following Phx actually recapitulates the pattern formed as part of embryonic liver development.

Discussion

DNA Methylation Programming.

DNA methylation patterns constitute an important layer of information that plays a key role in cell programming and development. Unlike other epigenetic marks, such as histone modification, methyl groups are covalently bound to DNA and, as a result, constitute an especially stable effector for molding chromatin structure and accessibility. Thus, both during development as well as in individual differentiated cell types, this layer of methylation plays a key role in cell programming, and, when the enzymatic machinery necessary for setting up these profiles is deleted, proper cell differentiation and development do not take place (12, 15, 18, 19).

Mechanism of Regeneration.

In this study, we have extended our understanding of this basic process by asking what happens to epigenetic programming during liver regeneration. It is well documented that following Phx, the remaining hepatocytes undergo expression changes that allow them to rapidly proliferate and, in this way, regenerate liver mass, but the underlying program directing these processes has not been elucidated. On the one hand, the transformation from being a nonreplicating stable cell type to being able to reenter the cell cycle and bring about proliferation could be carried out through TF–mediated changes in expression on a transient basis (20). Alternatively, this process could entail genomic reprogramming of the DNA template by altering the methylation pattern. Our results strongly suggest that this process indeed requires demethylation of critical gene regulatory sequences. They also reveal a unique perspective on how DNA methylation is managed during development.

Role of DNA Methylation in Liver Regeneration.

Seven days following standard Phx, we extracted the regenerated liver and isolated hepatocytes. These cells are presumably at the tail end of their proliferative stage (3, 21–23) and thus different from normal mature hepatocytes. After carrying out genome-wide DNA methylation analysis, we used bioinformatics to identify sites that have a different pattern than what is seen in normal hepatocytes and confirmed that these defined elements represent gene regulatory regions, such as enhancers. Furthermore, both motif and nearest-gene analysis indicate that these sites are involved in the control of embryonic development, including both cell proliferation and differentiation, and this has been confirmed by identifying specific expression changes associated with these methylation events. Finally, we performed a highly specific deletion of the machinery necessary to carry out the biochemical reactions that mediate removal of methyl groups and demonstrated a dramatic decrease in the proliferation marker Ki67, as well as inhibition of liver regeneration itself, thus providing proof for the concept that the demethylation induced by hepatectomy is required for proper regeneration. These results are consistent with the idea that hepatectomy induces regeneration by actually causing a basic reprogramming of the DNA template, perhaps in a manner similar to what occurs during earlier stages of hepatocyte development.

Model for Liver Development and Repair.

In order to test this idea, we first compared the expression profile of cells following Phx to that seen in earlier stages of hepatocyte development. Previous studies have already identified a number of developmental genes that appear to be reactivated during liver regeneration (24). Strikingly, however, a comprehensive RNA-seq analysis revealed that over 70% of genes induced during Phx had been previously activated during the early stages of liver development, clearly linking these two processes. Analysis of DNA methylation during these early stages indicated that while the sites that undergo demethylation during Phx were still methylated in early 11.5 d hepatoblasts, they then became specifically demethylated in later-stage embryonic cells (~E16.5), remaining this way until birth when they then undergo remethylation as the liver differentiates into mature hepatocytes (Fig. 6).

Fig. 6.

Developmental dynamics of Phx-induced methylation changes. This diagram shows the developmentally dynamic methylation state (methylated, yellow; unmethylated, blue) for regulatory regions that undergo demethylation (−, green bar) or de novo methylation (+, purple bar) following Phx, as well as those that undergo stable tissue-specific demethylation in hepatocytes (TS, red bar). Note that all of these regions are originally de novo methylated at the time of implantation. Tissue-specific sites undergo demethylation with the formation of terminally differentiated hepatocytes and remain in this form for the life of the organism. In contrast, the changes that occur following Phx represent a reversal to their methylation pattern seen in midembryo hepatoblasts (~E16.5 d). These same sites then go through normal epigenetic differentiation as the newly generated hepatocytes undergo maturation. Thus, the methylation changes that occur following Phx constitute a program of temporary and only partial dedifferentiation. This is in contrast to the tissue-specific elements which are programmed to remain stably demethylated, even in the event of injury.

According to this idea, the changes seen as a result of hepatectomy reflect an epigenetic reversal to an earlier proliferative-stage cell type that may have been responsible for the increase in liver mass during normal embryonic development. These cells are originally formed from early hepatoblasts by bringing about the programmed demethylation of the 1,684 regulatory regions identified in our studies, thus triggering an expression profile that drives early hepatic proliferation and differentiation. Once this stage has been completed, these sites undergo remethylation, while a different set of 435 regulatory regions, often associated with the same genes (25), undergo demethylation, perhaps influencing their associated genes to advance to the next stage of liver maturation. In this model, midembryonic hepatoblasts (12.5 to 16.5 d) only exist as a developmentally transient progenitor population, which is automatically eliminated as a result of epigenetic closure, the onset of terminal differentiation, and ultimate conversion of these cells to stable hepatocytes.

We suggest that liver injury signals a reversal of the now-closed core epigenetic proliferation program in all remaining hepatocytes, thus initiating liver regeneration. Once this proliferative response has been completed about 7 d after hepatectomy, this program is then turned off by returning the methylation pattern to its adult nonproliferative state (Fig. 1). Indeed, the very fact that we observe a multistep pattern of epigenetic changes during the course of liver regeneration strongly indicates that such a program already exists and is utilized during normal development. This picture is probably very similar to what occurs, for example, in the intestine where stem cells in the crypt produce transit-amplifying cells that proliferate and climb toward the villus where they simultaneously turn off their proliferative abilities and complete the last stages of differentiation to generate mature epithelial cells (26).

Methylation as a Memory Mechanism.

When taken together, our studies suggest that there are two types of methylation changes programmed into the genome, tissue specific and developmental stage specific. Tissue-specific changes occur during cell-type differentiation and are involved in defining the stable phenotype of these cells. These methylation patterns appear to be preserved throughout the life of the organism (12). Stage-specific methylation changes, on the other hand, take place during early embryogenesis and, rather than defining cell-type specificity, are probably involved in programming development-specific processes. These methylation events are evidently reversible, at first enabling cells to carry out their progenitor roles but then turning them off through epigenetic reversal in order to allow differentiation and lineage progression. The results of our paper suggest that it is this epigenetic plasticity that may make possible liver regeneration following Phx, completely independently of hepatocyte tissue-specific stability (Fig. 6).

DNA Methylation Dynamic Programming during Development.

The programming of DNA methylation during development is usually thought of in terms of the Waddington concept. In this model, the “epigenetic” changes that occur in the embryo make it possible to proceed forward along a development axis of increasing cell specificity and decreasing plasticity. At the level of DNA methylation, this would suggest that once methylation patterns are altered, they remain stably in place and cannot revert to an earlier stage or jump to a different parallel cell type. In general, this has proven to be the case for tissue-specific demethylation of regulatory regions since, once formed, they remain in their undermethylated state throughout the life of the organism, even after the initial demethylation machinery has been inactivated. This is, of course, made possible by the existence of a basic built-in mechanism for autonomously maintaining all DNA methylation patterns through cell replication.

Our studies bring to light a different form of epigenetic programming where certain regulatory sites are targets for demethylation at one stage of development but are also programmed to be able to undergo remethylation at a later stage. In the case of tissue regeneration, it appears to be possible to reinitiate the earlier demethylation program. On the other hand, regulatory sites involved in terminal tissue differentiation are not programmed to be reversible and, as a result, remain in their undermethylated state, even as the cells recapitulate an earlier developmental process and acquire properties of progenitors.

Reprogramming in Medicine.

These results have clear-cut implications regarding the use of reprogramming for tissue replacement by regeneration in man. As has been previously reported, dedifferentiation of mature cell types constitutes a highly unfavorable reaction. Our studies provide strong support for the idea that it is the DNA methylation profiles that are responsible for this irreversibility, mainly by strictly regulating cell propagation. However, once more is known about the various factors that mediate the normal developmental induction of specific methylation changes involved in forward and reverse differentiation, it should then be possible to devise new molecular strategies for effective and stable regeneration in many tissues of the body.

Materials and Methods

Animal Studies.

All animal experiments were performed in accordance with the guidelines of the Hebrew University Institutional Committee for the Use of Animals for Research. Female C57BL mice aged 10 to 12 wk were maintained in a Specific Pathogen-Free animal facility (Hebrew University-Hadassah Medical School Jerusalem, Israel), housed under regular 12 h light–dark cycle. Tet2/Tet3-deletion mice were generated by injecting 1,000 genome-equivalent copies of AAV8 carrying the gene for Cre under the hepatocyte-specific Tbg promoter (Penn vector) into the jugular vein of 3-mo-old Tet2^F/Tet3^F mice. Phx was carried out on 12 wk old mice by removing 2/3 of the liver, including resection of the left lower, upper lobes and the right upper lobe. Mice were anesthetized and subjected to liver perfusion. The whole liver was then digested by collagenase (Roche), and hepatocytes were removed and isolated by means of a Percoll gradient. Cells were then sorted by FACS according to size. This yielded a population of relatively pure hepatocytes.

Hepatoblasts were prepared from E16.5 mouse embryos by collecting fetal liver samples from littermate siblings which were then minced and digested for 30 min at 37 °C with 0.025% collagenase IV (Sigma-Aldrich, C4-22-1G). After gentle pipetting, the cell suspension was filtered through a 40 µm cell strainer and washed twice with RPMI medium containing 10% FBS. The cells were then stained for 40 min at 4 °C with moderate agitation with Dlk1-APC-conjugated mouse Pref-1/DLK1/FA1 antibody (R&D Systems, FAB8634A-100). Dlk1-positive hepatoblast cells were then isolated by flow cytometry.

DNA Methylation Analysis.

DNA was isolated from purified hepatocytes and incubated in lysis buffer [25 mM Tris-HCl (pH 8), 2 mM EDTA, 0.2% SDS, and 200 mM NaCl] supplemented with 300 μg/mL proteinase K (Roche) followed by phenol:chloroform extraction and ethanol precipitation, and RRBS libraries were prepared (8) and run on HiSeq 2500 (Illumina) using 100 bp paired-end sequencing. DNA methylation was analyzed by using sequencing reads from RRBS that were trimmed and quality filtered by trim galore software (v0.6.7), fastQC (27), and cutadapt (28), using default parameters for RRBS. Read alignment (genome build mm10) and extraction of single-base resolution methylation levels were carried out by BISMARK v0.23.1 (29). Percent methylation was calculated for one-hundred-base-pair tiles with a minimum coverage of 10 CpGs (25).

Differential methylation analysis was performed on treated and control samples with the R methylKit package v1.18.0 (30). CpG sites, tiled using a window and step size of 100, were filtered for minimum coverage depth of 15 for each of the treated and control samples. A set of tiles was selected to include at least four treated and four control replicate samples (custom R script). Tiles exhibiting a minimum methylation difference of 25% (averaged over samples) and minimum q-value of 0.01 were selected as the set of DMRs. We calculated, for each individual sample, the percentage of tiles that conform with the strict methylation difference and verified that nonconforming tiles are a minority for each of the samples. Methylation data from implantation stage embryos (GSE34864), 1 to 3 wk and adult hepatocytes (GSE85251) was analyzed using the same pipeline that was employed for our own data. Motif analysis was carried out by HOMER with an RRBS background (31), which is available online at http://homer.ucsd.edu/homer. For pathway enrichment analysis, we used STRING.

ATAC-seq Analysis.

ATAC-seq paired-end fastq files were downloaded from GSE158865, filtered, and trimmed with trim_galore ver. 0.6.1. Processed files were aligned to the mm10 assembly with hisat2 (32) and converted to bigwig files with deepTools (33) bamCoverage function. Treated samples (48 h) are compared with untreated samples (0 h) using deepTools bamCompare to obtain the log2 ratio between the samples following individual scaling by read counts. Distribution histograms are created with custom R script by counting scores of the ratio coverage file across hypomethylated DMRs (1,684 tiles). For the control, ATAC data are tested against a set of random 1,684 tiles, which are nonetheless methylated by 70% or more.

RNA-seq Analysis.

Raw reads (fastq files) were processed prior to alignment to remove bad-quality bases (from both ends) and adapter sequences. Adapter removal was done with cutadapt (version 1.7.1), using a minimal overlap of 1 and filtering out reads that became shorter than 15 nt. Following trimming, bad-quality reads were filtered out with the fastq_quality_filter program of the FASTX package (version 0.0.13), using parameters –q 20 and –p 90. Processed reads were aligned to the mouse genome with TopHat (version 2.0.13), allowing up to three mismatches per read. The genome version was GRCm38, with annotations from Ensembl release 84. Quantification was carried out with the Cufflinks package (version 2.2.1). The cuffquant program was used with the genome bias correction, the multimapped read assignment algorithm, and with masking for genes of untrusted signal, such as Ribosomal RNA (rRNA), MicroRNA (miRNA), Immuglobulin (IG), and T-Cell Receptor (TCR) loci. The cuffquant output was then used to calculate raw counts for each gene with the cuffnorm program.

Normalization and differential expression were done with the DESeq2 package (version 1.10.1). Genes with a sum of counts less than 2 over all samples were filtered out prior to normalization, followed by dispersion and size factor calculations. Differential expression was calculated with default parameters. The significance threshold for the comparison of the four samples from pregnant mice (P) to the two samples from virgin mice (V) was taken as padj < 0.1.

All RNA-seq datasets used for this study can be found in the GEO database under accession numbers GSE125006 (hepatocytes) and GSE90047 (hepatoblasts). Raw data were downloaded from the web. The quality of reads was verified using FastQC. Adapter and quality trim were performed using cutadapt with the suitable options and a quality threshold of 25. Reads were aligned to the Mus_musculus GRCm38 genome and annotated with Ensembl.GRCm38.gtf by STAR aligner. Count tables were obtained using the featureCounts tool. Further analysis was done using RStudio and relevant packages (R version 3.5.2). Counts were normalized with the DESeq2 package, as described. Genes with an adjusted P-value of more than 0.1 were dropped off. Genes were considered differentially expressed when their absolute value of log2-fold change was more than 1. Gene names were assigned using biomaRT.

Staining: Immunohistochemistry (IHC) and immunofluorescence (IF).

All tissues were fixed for 24 h in 4% formaldehyde, embedded in paraffin (FFPE), and cut into 5 μm thickness sections. The whole liver (intact, 48 h or 7 d after Phx) was paraffinized and sectioned with 5 μm steps. Slides were deparaffinized with xylene–ethanol gradient, steamed in 10 mM sodium citrate pH8 for 3 min at 120 °C for antigen retrieval, blocked in CAS-Block (ThermoFisher, cat. 008120) for 40 min, and then incubated in primary antibody diluted in CAS-Block overnight.

Antibodies were diluted 1:100 for SOX9 (Abcam, cat. ab185230) or 1:200 for KI67 (Abcam, cat. ab16667). The next day, slides were washed in PBS and incubated with DAB reagent (DAKO K4063), followed by a final incubation in PBS and dH₂O before mounting the coverslip in Permount (Fisher Scientific, SP15-500).

In order to quantify hepatocyte square area, liver slides were stained with anti-β-CATENIN (Abcam, cat. ab11576) as a primary antibody (diluted 1:100) overnight at 4 °C and Alexa Fluor 488 (Invitrogen, cat. A11029, 1:2,000) as a secondary antibody for 2 h at room temperature. Slides were covered with DAPI-containing media (Abcam, cat. ab104139).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

FASTQC files data have been deposited in GSE224133 (34). All other data are included in the manuscript and/or SI Appendix.