Stepwise reprogramming of liver cells to a pancreas progenitor state by the transcriptional regulator Tgif2

Abstract

The development of a successful lineage reprogramming strategy of liver to pancreas holds promises for the treatment and potential cure of diabetes. The liver is an ideal tissue source for generating pancreatic cells, because of its close developmental origin with the pancreas and its regenerative ability. Yet, the molecular bases of hepatic and pancreatic cellular plasticity are still poorly understood. Here, we report that the TALE homeoprotein TGIF2 acts as a developmental regulator of the pancreas versus liver fate decision and is sufficient to elicit liver-to-pancreas fate conversion both ex vivo and in vivo. Hepatocytes expressing Tgif2 undergo extensive transcriptional remodelling, which represses the original hepatic identity and, over time, induces a pancreatic progenitor-like phenotype. Consistently, in vivo forced expression of Tgif2 activates pancreatic progenitor genes in adult mouse hepatocytes. This study uncovers the reprogramming activity of TGIF2 and suggests a stepwise reprogramming paradigm, whereby a 'lineage-restricted' dedifferentiation step precedes the identity switch.

Figure 1. TGIF2 controls pancreatic and hepatic cell lineage divergence.

(a) RT-qPCR analysis of Tgif2 expression in the mouse foregut (fg) endoderm and its derivatives, liver and pancreas. Data were normalized to that of Sdha and represented as fold change (FC) compared with liver samples (set to 1). E8.5 fg was compared with E10.5 liver sample. Values shown are mean±s.e.m. (n=3) *P<0.05. (b) TGIF2 (red) colocalizes with FOXA2 (green) in the E7.5 mouse anterior DE and with PROX1 (green) in the E8.5 at the fg lip (see arrowhead). Arrow indicates prospective hepatic endoderm (HE). Inset panels depict TGIF2 single channel image of the area in the dashed box. Embryos are presented in lateral view. Scale bars, 50 μm. ht, heart; vpa, ventral pancreas. (c) Left, whole-mount in situ hybridisation analysis of Tgif2 in 2S-stage mouse embryo. Embryo is presented in ventral view; arrow indicates lateral domains of the ventral fg. Right, in situ hybridisation on E12.5 mouse cryosections detects expression of Tgif2 in the whole pancreatic epithelium (demarcated by yellow dotted line). Scale bar, 50 μm. pa, pancreas; st, stomach. (d) Schematic showing directed differentiation of mESC cultures into DE and, subsequently, pancreatic (PE) or hepatic endoderm (HE). On day (d) 8 of differentiation, PE and HE populations were analysed by RT-qPCR for the expression of the indicated genes in the presence or absence of LV-TGIF2. Induction of endogenous Tgif2 transcript was observed during directed differentiation of ESCs toward endoderm, consistent with its expression profile in the gastrulating mouse embryo (b). Data were normalized to that of Sdha and shown as Log2-expression ratio relative to control undifferentiated ES cultures. Serp., Serpina1. Values shown are mean±s.e.m. (n=3). *P<0.05, **P<0.01 obtained with REST randomization test for differentiated PE or HE versus undifferentiated ESC; ^#P<0.05, ^##P<0.01 obtained for PE or HE+LV-TGIF2 versus non-transduced PE or HE.

Figure 2. The TALE Tgif gene family controls lineage segregation within the ventral foregut endoderm of the mouse embryo.

(a) Representative maximum confocal z-projections of PDX1/SOX17 and PDX1/SOX17/PROX1 IF analyses in control (CTRL), Tgif1^−/−;Tgif2^lox/+;Sox2-Cre⁺ (referred to as T1−/−; T2r/+) and Tgif1^+/−;Tgif2^lox/lox;Sox2-Cre⁺ (referred to as T1+/−; T2r/r) mouse embryos. PDX1 (red) marked both dorsal and ventral pancreatic buds and at E9.5 co-localized with SOX17 (green) in the pancreatobiliary progenitor population that gives rise to the ventral pancreas, extrahepatic ducts, and gall bladder (arrow). PROX1 (blue) marked both liver and pancreatic buds. dpa, dorsal pancreas; lv, liver; r, recombined; vpa, ventral pancreas. Scale bars, 50 μm. (b) Quantification of the ventral pancreas and liver volume buds from confocal images of E9.5 embryos. Average fluorescence intensity of PROX1⁺ liver buds, SOX17⁺ and PDX1⁺ ventral pancreas buds was measured using the ‘object analysis' function in Huygens software. Error bars indicate±s.e.m. n, indicated for each genotype. Statistics by two-tailed t-test and P values are shown.

Figure 3. Tgif2 expression induces adult murine liver cells to acquire molecular features of pancreatic progenitors.

(a) Map of LVs and schematic of the ex vivo reprogramming. Murine adult primary or BAML hepatocytes (HEP) were transduced with the constitutively active PGK-TGIF2-EGFP LV (LV-TGIF2) expressing Tgif2 and EGFP, FACS-sorted for EGFP and characterized at different time points by various ex vivo and in vivo approaches. (b) RT-qPCR analysis of hepatic and pancreatic gene expression in mouse primary HEPs at d7 and d14 after transduction with LV-TGIF2. Data were normalized to Sdha and represented as Log2-expression ratio between LV-TGIF2-transduced and control hepatocytes at matched time-points. (c) RT-qPCR analysis of hepatic and pancreatic gene expression in murine adult BAML HEP cells transduced with LV-TGIF2 at the indicated time points after transduction. Data were normalized to Sdha and represented as Log2-expression ratio between LV-TGIF2-transduced and control cells. Values shown are mean±s.e.m. (n=5) *P<0.05. (d) IF of unsorted (no FACS) LV-TGIF2-transduced (right) BAML HEP cells and control cells (left). In top panel of LV-TGIF2-transduced cells, asterisks (*) indicate Albumin (ALB)-positive cells that are PDX1-negative; open arrowheads (>) indicate PDX1/GFP-positive cells that are ALB-negative in transduced cultures. In middle panels, arrows indicate examples of GFP-positive cells either PDX/SOX9-double or PDX1/TGIF2-double positive cells. In bottom panel, asterisks (*) indicate Glutamine synthetase (GS)-positive cells (blue) that are negative for PDX1 (red) in cultures d50 after transduction; open arrowheads (>) indicate PDX1-positive cells that are negative for GS. Scale bars, 50 μm.

Figure 4. Primary hepatocytes transduced with LV-TGIF2 formed pancreatic organoid structures.

(a) Micrographs of primary HEPs grown in three-dimension (3D) culture conditions for E10.5 mouse pancreas progenitor cells. Control (CTRL) primary hepatocytes failed to grow in the same conditions. All organoid structures shown in the figure were generated from LV-TGIF2-transduced primary hepatocytes. Bar, 100 μm. (b) IF staining on cryosections of LV-TGIF2-organoids for GFP (green), SOX9 (blue) and PDX1 (red). Hoechst (Hoe) nuclear counterstain in grey. Scale bar, 20 μm. (c) Whole-mount IF staining of pancreatic explants transplanted with LV-TGIF2-organoids. Insets are magnifications of the areas in the white boxes and shown as single channel images. Left panel, arrows indicate GFP-positive grafted cells that are positive for PDX1 and NKX6.1 (n=6, IF exp.). Middle panel, arrow indicates GFP-positive cell that is amylase-positive (n=5, IF exp.). Right panel, arrows indicate GFP-positive grafted cells expressing endocrine markers, insulin or glucagon (n=6, IF exp.). Dashed yellow line marks pancreatic epithelium. Scale bar, 50 μm.

Figure 5. Lineage conversion is accompanied by global transcriptional remodelling.

(a) Hierarchical clustering of the gene expression values of the sub-set of genes that were differentially expressed between BAML HEP control and TiPP cells at d14, derived from BAML HEPs+LV-TGIF2 (total 592 genes; P<0.01 and FC >4 or <−4) among all samples analysed. Gene probes are located in rows and samples in columns. Gene expression across rows in the heatmap are coloured according to the z-score, such that the mean expression of each row is set to white colour and expression values higher or lower than the mean are graded towards red or blue, respectively. Rows in the heatmap are ordered according to the FC. (b) Heat map (left) illustrating the relative expression levels of selected hepatic genes involved in metabolic activities in HEP control and TiPP cells. Heat map (right) illustrating the relative expression levels of selected pancreatic progenitor genes in HEP control and TiPP cells. The colours in the heatmap are normalized per row by calculating the z-score for each row. Replicates were summarized as the mean. (c) Panther pathways enrichment analysis of differentially regulated genes in HEP control and TiPP cells at d14 (P<0.05 and FC >2 or <−2). For complete gene ontology annotation lists see Supplementary Data 1. (d) Heat map illustrating the relative expression levels of selected Wnt signaling pathway genes. Genes in bold were selected for confirmation by RT-qPCR. (e) Confirmation of microarray gene expression changes by RT-qPCR analysis. Data were normalized to Sdha and represented as Log2-expression ratio between TiPP d14 cells derived from either BAML HEP or primary HEP and their respective control cells. Values shown are mean±s.e.m. (n=3) *P<0.05.

Figure 6. In vivo analysis of TGIF2 reprogramming activity.

(a) Transplantation assays in Akita diabetic mice. Reprogrammed TiPP cells (derived from BAML HEPs+LV-TGIF2 d8 post-transduction) were grafted under the KC of hyperglycemic Akita mice (red line). Animals engrafted with BAML HEP cells transduced with LV-GFP were used as negative controls (green line). Blood glucose levels were measured under non-fasting conditions before transplantation (time points: −2 wk and −1 wk), the day of transplantation (time point 0; arrow) and, subsequently, twice a week after transplantation. n=6 animals in each group. Normoglycemia is defined as blood glucose level below 200 mg dl⁻¹ in wild-type littermate mice under nonfasting conditions (dotted line). Values shown are mean±s.e.m. **P<0.01 (two-way ANOVA comparison test between LV-GFP and TiPP transplanted groups). (b) IF analysis on cryosections of mouse KCs transplanted with LV-GFP BAML HEP cells and TiPP cells. Hoechst (Hoe) nuclear counterstain in blue. Insets show area in the dashed box of: PDX1 (red)/GFP (green) staining without SOX9 channel (blue); C-peptide (C-PEP) (red)/Hoechst (blue) staining without GFP channel (green); PC2 (red)/Hoechst (blue) staining without GFP channel (green); UCN3 (red)/Hoechst (blue) staining without GFP channel (green); GLUT2 (red)/Hoechst (blue) staining without GFP channel (green), arrows indicate staining near the membrane, as expected for the location of mature Glut2 protein. To a lesser extent, some cells displayed Glut2 staining in the cytoplasm, which is similar to neonatal beta-cells. Scale bars, 50 μm.

Figure 7. In vivo analysis of TGIF2 reprogramming activity.

(a) Schematic of the AAV-mediated experimental strategy to express Tgif2 in vivo in the adult mouse liver. AAV-injected and uninjected control livers were examined at the indicated time points (*) after iv injection. (b) RT-qPCR analysis of indicated genes in AAV.TGIF2-injected livers at d30. Data were normalized to 36B4 and represented as Log2-expression ratio between AAV.TGIF2-injected and AAV.GFP-injected livers. Values shown are mean±s.e.m. (n=3). (c) IF staining of AAV-injected (+AAV.TGIF2) and uninjected (−AAV) adult mouse livers for GS, SOX9, Epcam, PDX1 and NKX6.1. Micrographs show IF stainings performed at d30 after AAV.TGIF2 injection; SOX9/NKX6.1 IF staining was done at d60. SOX9 and Epcam mark biliary epithelial cells (bec) and are absent in hepatocytes in uninjected livers. Arrows indicate examples of SOX9−, PDX1− or NKX6.1-positive hepatocytes in AAV.TGIF2-injected livers. Outlined areas are shown in the insets without Hoechst nuclear counterstain (blue). CV, central vein. Scale bars, 50 μm. (d) Quantification of SOX9-positive hepatocytes and PDX1-positive cells in adult livers injected with AAV.GFP (n=2 d30), AAV.TGIF2 (n=3 d7; n=4 d30) and uninjected (n=3). Average number of labelled cells per section (mm²) was determined for each animal in one entire liver lobe. Results are expressed as the mean±s.e.m. Statistics by two-tailed t-test. **P<0.01.