Cellular reprogramming in vivo initiated by SOX4 pioneer factor activity

Abstract

Tissue damage elicits cell fate switching through a process called metaplasia, but how the starting cell fate is silenced and the new cell fate is activated has not been investigated in animals. In cell culture, pioneer transcription factors mediate "reprogramming" by opening new chromatin sites for expression that can attract transcription factors from the starting cell's enhancers. Here we report that SOX4 is sufficient to initiate hepatobiliary metaplasia in the adult mouse liver, closely mimicking metaplasia initiated by toxic damage to the liver. In lineage-traced cells, we assessed the timing of SOX4-mediated opening of enhancer chromatin versus enhancer decommissioning. Initially, SOX4 directly binds to and closes hepatocyte regulatory sequences via an overlapping motif with HNF4A, a hepatocyte master regulatory transcription factor. Subsequently, SOX4 exerts pioneer factor activity to open biliary regulatory sequences. The results delineate a hierarchy by which gene networks become reprogrammed under physiological conditions, providing deeper insight into the basis for cell fate transitions in animals.

Fig. 1. Sox4 and Sox9 are required for biliary reprogramming.

a Flow cytometry for CD24 and EPCAM at the indicated time points (weeks) following DDC challenge (n = 4, 4, 15, and 21 animals for 0, 4, 8–9, and 12–13 wpi, respectively). b Schematic showing the strategy for isolating early reprogrammed cells (EGFP+CD24− EPCAM−), early-to-intermediately reprogrammed cells (EGFP+CD24+EPCAM−), and intermediate-to-late reprogrammed cells (EGFP+CD24+EPCAM+). c PCA of whole transcriptomes (RNA-Seq) for the indicated cell populations along the hepatocyte-to-biliary axis (n = 3 animals). d Results of RNA-Seq showing the expression of Sox4 and Sox9 during DDC-induced biliary reprogramming. Expression levels are normalized by TPM (transcripts per million) (n = 3 animals). e Schematic for the design of the loss-of-function experiment. AAV8 packaged with three sgRNAs targeting Sox4 and/or Sox9, and Hnf1b along with TBG-Cre was injected to Rosa26-LSL-Cas9-EGFP mice to induce knockout, while enabling the simultaneous genetic labeling of the infected hepatocytes. Biliary reprogramming was then induced by introducing a DDC diet. Animals were sacrificed for analysis 9 weeks later. f Reprogramming efficiency was assessed by flow cytometry using two surface markers, CD24 and EPCAM, which serve as surrogates for early and late reprogrammed cells, respectively (n = 14, 10, 10, 10, and 5 animals for empty vector, Sox4 KO, Sox9 KO, Sox4/9 DKO, and Hnf1b KO conditions, respectively). Exact P values by two-sided t-test vs. the empty vector (EV) group as reference are shown without adjustment. g PCA mapping of the RNA-Seq data of the sorted EGFP+ cells (n = 4 animals). h GSEA for the RNA-Seq data using hepatocyte-enriched and reprogrammed cell-enriched signatures (n = 4 animals). The signature gene sets were curated using previously published RNA-Seq data by comparing the gene expression of normal hepatocytes and DDC-induced reprogrammed cells (YFP+EPCAM+ cells). Adjusted P values and NES values were calculated using the R fgsea package. (a, d, f) The center line, box limits, whiskers, and points indicate the median, 25th/75th quartiles and 1.5× interquartile range, respectively.

Fig. 2. Ectopic expression of Sox4 is sufficient to induce the early stage of biliary phenotypes in adult hepatocytes.

a Experimental design. AAV8 packaged as shown was injected into Rosa-LSL-Cas9-EGFP mice. Total liver cells or enriched hepatocytes were harvested after 2 weeks on normal diet. b qRT-PCR of Sox4 and Sox9. c IF of SOX4 and SOX9 at the designated time points. Arrows indicate nuclear staining of SOX4 or SOX9 in EGFP+ hepatocyte-derived cells. Arrowhead indicates SOX9 staining in BECs. d Reprogramming efficiency assessed by flow cytometry of CD24 and EPCAM. Exact P values by two-sided t-test with the empty vector (EV) group as reference are shown without adjustment. e Heatmap of qRT-PCR for biliary and hepatocyte genes. Expression levels are normalized to Actb and shown as z-scores for each gene. f Kinetics of Sox4 mRNA expression assessed by qRT-PCR (n = 2, 2, 2, 3, and 2 animals for Sox4 samples at 1, 2, 4, 7, and 10 dpi, respectively). Data for the EV control pooled from 4, 7, and 10 dpi (n = 4 animals). g H&E of AAV-HA-Sox4-injected or control livers at designated time points. Arrows indicate atypical ductal cells extending from portal vein (PV) regions. h Representative IF of reprogramming markers in livers harvested from EV- or Sox4-injected mice at the designated timepoints. Arrows indicate EGFP+ hepatocyte-derived cells expressing the indicated marker. i PCA plot of RNA-Seq data from EV- and Sox4-injected mice compared to reprogrammed cells following DDC treatment (n = 3 animals). j GSEA comparing EV and Sox4 hepatocytes, using the hepatocyte-enriched and early reprogrammed cell (Rep_early)-enriched signatures (n = 3 animals). Adjusted P values and NES values were calculated using the R fgsea package. b, f Expression normalized to Actb, with EV hepatocyte expression set to one. b, d, e: n = 4, 3, 5, and 3 animals for EV, Sox4, Sox9 and Sox4/9 conditions, respectively. b, d, f The center line, box limits, whiskers, and points indicate the median, 25th/75th quartiles and 1.5× interquartile range, respectively. c, g, h Scale bar = 50 μm. Staining experiments used samples from at least 2 animals.

Fig. 3. The chromatin landscape following ectopic expression of Sox4 recapitulates that of partially reprogrammed cells.

a Differential peak analysis at 4 dpi showing newly opened, newly closed, and unchanged regions following Sox4 expression in hepatocytes compared with empty vector (EV). b Averaged aggregate plots of the ATAC-Seq signals in the newly opened, unchanged, and closed regions corresponding to panel a. c Schematic describing ATAC-Seq-based footprinting as implemented by TOBIAS. d Results of DNA footprinting analysis for the SOX4 binding motif comparing empty vector and Sox4 hepatocytes at newly opened and unchanged regions as defined in (a). e PCA mapping for ATAC-Seq of empty vector EV and Sox4 hepatocytes (n = 3 animals) at 4 dpi. Data obtained for this study were downsampled by approximately 1/100-fold to adjust read depth and compared with previously published data^,. “Rep_intermed” indicates SOX9+ cells sorted from Sox9-RFP reporter mice treated with DDC. f Results of HOMER motif analysis using all SOX4 peaks from CUT&RUN-Seq (n = 9,463 at 18 hpi; n = 19,362 peaks at 4 dpi). The top 15 motifs from each timepoint ranked as p-values are shown. SOX motifs, highlighted in red, were highly ranked. The HNF4A motif, highlighted in green, was the third most significantly enriched motif at both 18 hpi and 4 dpi. g Quantification of SOX4 total binding was estimated using E. coli spike-in genomic DNA for scaling. Data are shown as fold-change compared to mean values at 18 hpi data (n = 3 animals). The center line, box limits, whiskers, and points indicate the median, 25th/75th quartiles and 1.5× interquartile range, respectively. h Averaged aggregate plots for spike-in-scaled SOX4 CUT&RUN-Seq data at newly closed and newly opened regions at the designated time. i SOX4 CUT&RUN signals (right two columns) visualized as heatmaps for the newly opened and newly closed regions defined by ATAC-Seq (left three columns). Corresponding averaged aggregate plots for SOX4 CUT&RUN are shown in Supplementary Fig. 11b. Note that SOX4 data are normalized genome-wide in each sample (18 hpi and 4 dpi) and do not support quantitative comparison between the two timepoints.

Fig. 4. SOX4 binding silences hepatocyte genes while priming biliary genes by altering binding patterns.

a GSEA was performed with genes annotated to 4dpi ATAC-Seq peaks to compare empty vector and Sox4 hepatocytes using the hepatocyte-enriched and Rep_early-enriched signatures (see Supplementary Data 1, Fig. 2j, Methods). Heatmap visualization of the 5,272 genes near newly closed peaks (b) and 5,823 genes near newly opened peaks c using the RNA-Seq data of the DDC-induced reprogramming experiment (Fig. 2i, j) (n = 3 animals). Genes were categorized to four (b) or three (c) clusters by k-means clustering. Representative hepatocyte (39/50) and biliary/reprogramming genes (33/49 from a manually curated gene list; Supplementary Table 1) are shown on the right. Newly closed (d) and opened peaks (f) in Sox4 expressing hepatocytes at 4 dpi compared with empty vector hepatocytes (Fig. 3a) were annotated with the nearest genes, and this gene list was used as the input for GO enrichment analysis (n = 3 animals). GO analysis of RNA-Seq was also performed using genes downregulated (e) and upregulated (g) in Rep_intermediate cells vs Hep during DDC-induced reprogramming (n = 3 animals). The same analysis comparing different stages, namely Rep_early vs Hep and Rep_late vs Hep are also shown in Supplementary Fig 12a, b. GO terms shared between the newly-closed and newly opened region-associated gene set and DDC-induced reprogramming context at any reprogramming stages are highlighted in bold blue (d, e) and red (f, g) texts, respectively. h GSEA was performed with genes annotated to the SOX4 CUT&RUN peaks identified either at 18 hpi and 4 dpi using the hepatocyte-enriched and Rep_early-enriched signatures as described earlier (Supplementary Data 1, Fig. 2j, Methods). i Global footprint analysis depicted as a volcano plot. The analysis used all motifs assigned as “bound” by TOBIAS in either empty vector or Sox4 hepatocytes (n = 3 animals). Differential footprints are defined as those with Log2(fold-change of footprint score) >0.15 and log10(p value) < −100. Statistical testing was performed using the TOBIAS tool with the default setting. a, h Adjusted P values and NES values were calculated using the R fgsea package.

Fig. 5. SOX4 suppresses hepatocyte identity by inactivating hepatocyte enhancers.

a Newly closed regions with or without overlaps with 18hpi-SOX4 peaks were annotated with their nearest genes, and expression of each gene in empty vector or Sox4-expressing hepatocytes was compared with the DDC-induced reprogrammed cells. P values were calculated by Wilcoxon rank sum test with Hep_EV as the reference are shown without adjustment (n = 3 animals). The center line, box limits, whiskers, and points indicate the median, 25th/75th quartiles and 1.5× interquartile range, respectively. b Schematic representing the nucleosomal states and the corresponding outcomes of MNase-Seq nucleosomal states are roughly categorized into four groups, with representation of how they are detected by different levels of MNase treatment. c Heatmap representation of MNase-Seq at low and high levels for the newly closed regions with or without overlaps with SOX4 peaks at 18 hpi along with active liver promoters and enhancers. The corresponding averaged aggregate plots are shown in (d). d Averaged aggregate plots of hepatocyte MNase-Seq in newly closed regions with or without overlaps with SOX4 peaks (18 hpi) along with active liver promoters and active liver enhancers. e Heatmap representation of ChIP-Seq for core histone H2B and H3 for the same regions as in (c). The corresponding averaged aggregate plots are shown in (f). f Averaged aggregate plots of H2B and H3 ChIP-Seq for the same regions as in c. g ATAC-Seq and CUT&RUN-Seq of SOX4, H3K27ac, H3K4me1 and H3K27me3 signals visualized as heatmaps at active liver enhancer regions. The corresponding averaged aggregate plots are shown on the top. Aggregate plots with y-axis values are shown in the Source Data File. h Genome browser view of the Cyp2e1 gene is shown as an example of SOX4-induced closing of a hepatocyte gene-associated region. HNF4A ChIP-Seq data were obtained from GEO datasets (GSE137066) reflecting normal liver cells. c–f Analyses done using GEO dataset GSE57559.

Fig. 6. SOX4 evicts hepatocyte transcription factors by hijacking their binding motifs.

a. HOMER motif analysis with known motifs for 18hpi- and 4dpi- enriched SOX4 CUT&RUN peaks. 18hpi-enriched SOX4 peaks (n = 2327) and 4dpi-enriched SOX4 peaks (n = 3136) were identified using DiffBind and DESeq2 packages (n = 3 animals). The HNF4A motif, highlighted in red, ranked first among 18 hpi-enriched SOX4 binding sites. b Consensus binding sequences of several hepatocyte transcription factors exhibiting partial overlap with that of SOX4 (overlapped sequences are underlined). c ATAC-Seq footprinting analysis of hepatocyte transcription factors HNF4A and RXRA at motifs overlapping the hepatocyte enhancers. The data are shown as the comparison between empty vector and Sox4 hepatocytes (n = 3 animals). Motifs assigned as “bound” by TOBIAS for the empty vector and Sox4 expressing samples are combined and used for the analysis. d Schematic representing the Cyp2e1 distal enhancer with an HNF4A motif for which a ds-DNA oligonucleotide and its scrambled mutant were designed. Most conserved sequence in the HNF4A motif is highlighted in red texts along with the consensus HNF4A motif sequence shown below. e EMSA was performed to assess the ability of recombinant mouse HNF4A protein and mouse SOX4 protein to bind the wild type and mutant Cyp2e1 enhancer sequence. f EMSA was performed to assess whether SOX4 can displace HNF4A pre-bound to the wild type Cyp2e1 enhancer dsDNA oligonucleotide by titrating SOX4 protein concentration used for competition. e, f A representative gel image from 2 replicates is shown.

Fig. 7. SOX4 opens putative biliary cell enhancers in hepatocytes.

a EMSA was performed to evaluate the binding ability of recombinant mouse SOX4 to naked LIN28B DNA and in vitro assembled nucleosomal LIN28B DNA. A representative gel image from 2 repeated experiments is shown. b SOX4 CUT&RUN-Seq data (right column) are shown for the more opened regions (MOR) and de novo opened regions (DOR), which correspond to newly opened regions with weak ATAC peaks in empty vector and those without ATAC peaks in empty vector (left two columns). c Heatmaps of previously published MNase-Seq data (GSE57559) obtained for adult hepatocytes at low and high levels of MNase. The regions are centered with all the intersectable ATAC peaks of either empty vector or Sox4-expressing hepatocytes. d Averaged aggregate plots corresponding to the heatmaps shown in c. e Heatmaps of previously published H2B and H3 ChIP-Seq obtained for adult hepatocytes (GSE57559). Regions are centered with all intersectable ATAC peaks of either empty vector or Sox4 hepatocytes. f Averaged aggregate plots corresponding to the heatmaps shown in (e). g Genome browser views of two examples for Sox4-induced opening at biliary epithelial cell genes: Itga3 (intermediate-to-late reprogramming marker) and Cd24a (early-to-intermediate reprogramming marker). h Genome browser views of ATAC-Seq and CUT&Tag-Seq data in the same regions as g. ATAC-Seq data of reprogrammed cells and biliary cells are adapted from our earlier study. Hep ATAC-Seq data indicates the Hep_EV data collected in this study, which were downsampled to adjust the read depth comparable to the DDC-induced reprogrammed and biliary cells (as shown in Fig. 3). CUT&Tag-Seq data were collected for each stage of DDC-induced reprogramming.

Fig. 8. Proposed model of the SOX4-mediated regulatory mechanism of hepatobiliary reprogramming.

Liver injury activates signaling pathways which are responsible for induction of biliary reprogramming, such as YAP, Notch and TGFβ pathways. Sox4 expression is readily activated as one of the earliest responses to these pathways. Upon translation, SOX4 binds hepatocyte enhancers through recognition of the shared motif sequences, and evicts the originally bound hepatocyte transcription factors, including HNF4A and RXRA, which leads to repression of hepatocyte phenotypes. Then, SOX4 starts to bind closed and silenced chromatin regions required for acquisition of biliary phenotypes and open these regions as a traditional pioneer factor to establish access for other factors to these regions.