Identification of a patchy hepatocyte proliferation pattern in Shp2-deficient (SKO) liver after PHx

(A and B) Immunofluorescence on liver tissue sections 2 days after PHx. HNF4α is a hepatocyte marker and Ki67 is a proliferation marker. Arrow in (A) points to an area enriched with proliferating hepatocytes. Dashed line in (B) shows an area with continuous CD133 expression.

(C) Quantification of proliferating rate in hepatocytes in CD133-positive and -negative areas in SKO livers 2 days after PHx. Each dot indicates one area. Data were collected from 3 mice. Means ± SEM are shown. ****P < 0.0001 (two-tailed unpaired t test).

(D) While WT hepatocytes proliferated at high frequency everywhere, proliferating hepatocytes in SKO liver were mostly located in patchy areas marked by CD133 expression.

(E) Immunofluorescence of CD133 on liver tissues at day 0 or 2, and 3 weeks (0d, 2d, 3 wk) after PHx. CD133+ hepatocyte clusters were only found in SKO livers after PHx (light green arrows). In WT livers, CD133 expression was only seen in bile duct epithelial cells (arrowheads).

Scale bars, 100 μm (A and E).

CD133+ colonies represent a unique regeneration process in Shp2-deficient liver

(A) Representative image of CD133+ colonies shown by immunofluorescence on SKO liver tissues 3 weeks after PHx. Dashed lines, vasculatures.

(B) H&E staining of SKO liver sections 3 weeks after PHx. Yellow dashed line: boundary between the colony and surrounding tissue; white dashed lines: vasculatures.

(C) Immunofluorescence on SKO liver sections at indicated time points after PHx. Arrowheads and dashed lines indicate the boundaries between the colonies and surrounding tissues, which was clear at 3 weeks but disappeared at 5 weeks.

(D) Unstained thick tissue section of SKO liver 3 weeks after PHx.

(E) Colonies (C) and non-colony control areas (N) were dissected out from thick tissue sections of SKO livers 3 weeks after PHx, and analyzed by immunoblotting. Random areas from WT livers 3 weeks after PHx, bile duct (BD) and E17.5 liver (E) were used for comparison.

(F) Illustration of epithelial lineages in the liver. Scale bars, 100 μm (A-D).

Mitogenic signal deficiency induced tight clustering of cells that leads to continuous growth

(A) Schematic illustration of clonal tracing performed in SKO livers during regeneration.

(B) Liver tissues were examined before (PHx 0d) or 3 weeks after the surgery (PHx 3wk). E-Cad shows the colony structures (arrows).

(C) GFP-labeled clone in a colony 3 weeks after PHx. Note that no unlabeled cells were detected on this edge (arrowheads).

(D) Cell numbers of GFP-labeled hepatocyte clones in colony and non-colony areas, counted on sections. Data were collected 3 weeks after PHx. Means ± SD are shown, n = 3, ****P < 0.0001 (Mann-Whitney test).

(E) Distribution analysis of clone sizes from (D).

(F) Clonal dynamics of colony-forming hepatocytes.

(G) Clone/colony size ratios to estimate the original colony size (see also Figure S2B). Mean ± SEM is shown.

(H) Immunofluorescence on the colonies (E-Cad+ clusters) in Shp2-deficient hepatocyte culture in vitro. Note the other non-colony cells with low proliferation rate, indicating the patchy proliferation in the colonies. See also Figure S2D.

Scale bars, 100 μm (B, C and H).

CD133 induction during signal deficiency is a widely conserved mechanism

(A and B) Immunofluorescence of liver sections 2 days after CCl4 injection. Proliferating hepatocytes were scattered in WT livers, whereas they were highly concentrated in CD133+ colonies (white dashed line) in SKO livers as shown in (A). CD133+/EpCAM+/HNF4α- bile duct epithelial cells (arrowheads) were not associated with the colonies. Pink dashed lines, injured areas. PV, portal vein.

(C) Immunofluorescence of Methep-/- liver sections 2 days after PHx or CCl4 injection. White dashed lines, CD133+ colonies. Pink dashed line, injured area.

(D) qRT-PCR analysis of PLC cell lysates treated with Shp2 or MEK inhibitors (Shp2i and MEKi). **P < 0.01, ***P < 0.001, ****P < 0.0001 (two-tailed unpaired t test, each compared with DMSO treatment). Means ± SD from 3 replicates are shown.

(E) Immunoblotting of PLC cell lysates treated with inhibitors or transfected with Shp2 targeting CRISPR vector. Guide RNA targeting the AAVS1 safe harbor site was used as a control (sgCtrl).

(F) qRT-PCR analysis of MCF10A cell lysates treated with the inhibitors. *P < 0.05, ***P < 0.001, ****P < 0.0001 (two-tailed unpaired t test, each compared with DMSO treatment). Means ± SD from 3 replicates are shown.

(G) Immunoblotting of MCF10A cell lysates treated with inhibitors.

(H) qRT-PCR analysis of lysates from various cell lines treated with Shp2 or MEK inhibitors. *P < 0.05, ***P < 0.001, ****P < 0.0001 (two-tailed unpaired t test, each compared with DMSO treatment). Means ± SD from 3 replicates are shown.

Scale bars, 100 μm (A-C).

Mitogenic signal deficiency induces CD133+ vesicles

(A and B) Immunofluorescence on in vitro colonies of primary hepatocytes from SKO liver. CD133 was localized at filament-like structures in E-Cad+ colonies as shown by arrowheads in (A), which were connected between different hepatocytes as shown by arrows in (B).

(C) 3D-reconstituted confocal image of immunofluorescence on PLC cells. Lower panel shows the Z-plane section of the orange box area. Arrowheads indicate the CD133 signal on continuous filament like structures bridged between neighboring cells. Pink dashed lines indicate the cell surface.

(D) Immunofluorescence of MCF10A cells treated with Shp2 inhibitor.

(E and F) Super-resolution STORM images of immunofluorescence on PLC cells without (E) or with (F) CD133 overexpression. Colocalization of CD133 and β-tubulin was analyzed as Pearson’s coefficient. Mismatched green and magenta channels from shuffled ROIs were measured as controls (Figure S4E). Means ± SEM from 6 images are shown. **P < 0.01, ***P < 0.001 (two-tailed unpaired t test).

(G) Immunofluorescence and Immuno-Gold EM images of cryo-ultramicrotome sections of SKO liver tissue after PHx. Cyan arrowheads and asterisk indicate apical lumens. White arrowheads indicate the CD133 signals aligned between apical lumens of neighboring cells. Light green arrow, CD133 staining (12 nm colloidal gold); Magenta arrows, α-tubulin staining (18 nm colloidal gold).

(H) Immunoblotting of CD133+ vesicles isolated from MEK inhibitor (MEKi) -treated PLC cells. Markers for different fractions were analyzed. CD133 antibody used for the vesicle isolation was IgG produced in mouse, which was detected by anti-mouse IgG antibody, showing efficient capture by the beads. Despite the efficient capture, the DMSO-treated PLC cells did not have much CD133+ vesicle to be bound with the antibody.

(I) EM image of Immunogold staining on the isolated vesicle.

Scale bars, 100 μm (A), 25 μm (D), 1 μm (E, F, Fluorescence in G), 100 nm (EM in G), 50 nm (I).

CD133+ vesicles contain mitogenic mRNAs and traffics between neighbor cells

(A) Agarose gel electrophoresis of total RNAs extracted from WT and SKO livers and CD133-positive vesicle and negative fractions from SKO liver after PHx. Arrowheads show rRNAs and arrow shows microRNAs.

(B) qRT-PCR analysis of RNAs extracted form WT (3 mice) and SKO tissues (3 mice) and from CD133+ vesicles and CD133− fractions (4 mice for both). Means ± SEM are shown. *P < 0.05, **P < 0.01 (uncorrected Dunn’s multiple comparison test, performed after Kruscal-Wallis test). n.s., not significant.

(C) RNA-seq analysis of the different cell types and the whole cells. The bars indicate proportions between numbers of deficient and enriched gene transcripts in each RNA types.

(D) Comparison of IEG contents between the different vesicle types with RNA-seq.

(E) RNA-FISH for MYC mRNA and immunostaining for CD133.

(F) Quantitative colocalization analysis of CD133 and MYC mRNA in PLC cells shown in (E). Means ± SEM from 14 images are shown.

(G) Experimental design to detect the traffic of CD133+ vesicles between neighbor cells.

(H) Immunostaining of Myc-tag and CD133 on GFP+ and mCherry+ PLC cells mixed as shown in (G). Note that Myc-tag only indicates exogenous CD133, while CD133 indicates both endogenous and exogenous CD133. Myc-tag was primarily detected in the GFP+ cells, but also detected on the bridges (arrowheads) and in the mCherry+ cells (arrows). GFP was not detected at the same locations (arrowheads and arrows), indicating the specific traffic of CD133-Myc-tag.

Scale bars, 10 μm (E) and 50 μm (H)

CD133-mediated mRNA sharing converts intercellular heterogeneity into intracellular diversity

(A) Immunoblotting of HuR in the CD133+ vesicles from SKO liver after PHx. GAPDH and HNF4α were used as controls for cytoplasmic and nuclear fractions, respectively.

(B) Immunofluorescence on PLC cells. Arrows, peri-nuclear areas enriched with HuR in the cytoplasm; Arrowheads show colocalization of HuR on the CD133+ filament bridging two cells.

(C) Immunofluorescence on PLC cells treated with the Shp2 inhibitor (SHP099). Arrowheads show strong localization of HuR on the CD133+ filaments.

(D) Immunofluorescence on PLC cells transfected with CD133 expression vector treated with a Shp2 inhibitor (SHP099). Arrowheads show strong localization of HuR on the CD133+ filaments.

(E) Immunoblotting of CD133+ vesicles isolated from MEK inhibitor-treated PLC cells.

(F) qRT-PCR analysis of PLC cell lysates after treatment with CD133+ vesicles isolated from MEK inhibitor-treated PLC cells. RNase and Triton X-100 were used to digest the RNA content of the vesicles. *P < 0.05 (two-tailed unpaired t test). n.s., not statistically significant. Means ± SD from 3 replicates are shown.

(G) A model and predictions for single cell RNA-seq data analysis.

(H) Total IEG expression levels. Cyclin D1-positive and -negative cells were analyzed separately to evaluate the influence of cell cycle on the IEG analysis. See Methods section for what the bars and dots represent.

(I and J) Box plots (Tukey’s) of IEG diversity within each cell (calculated as entropy) and IEG variations among cells. Analyses were focused on cyclin D1+ cells in all groups for fair comparison.

(K and L) Plot of intracellular IEG diversity against total IEG expression levels in SKO hepatocytes 2 days after PHx. Blue color gradient indicates cyclin D1 expression levels. For the simulation in (L), the parameters used were: Group of 5 cells, X=1/12, model 3 (see also Figure S8 and Methods). Green and Black arrows show typical profiles of CD133-positive and -negative cells, respectively.

(M) Box plot (Tukey’s) of intracellular IEG diversity after simulation. The analysis was not limited to cyclin D1-positive or -negative cells. Note the simulation of the IEG exchange attracted the cells from cyclin D1-low profile to cyclin D1-high profile (K-M). Statistics was performed by Wilcoxon rank sum test adjusted by FDR in (I), (J) and (M).

Scale bars, 50 μm (B, C), 5 μm (D).

CD133 is required for cell proliferation under proliferative signal deficit

(A and B) Immunofluorescence (A) on liver sections of WT and Prom1 KO mice 2 days after PHx with or without Shp2 deletion using AAV-Cre and quantification of Ki67+ ratio in hepatocytes in the indicated genotypes (B). Separate analyses of pericentral and periportal hepatocytes showed insignificance of zonal difference. **P < 0.01, (two-tailed unpaired t test). Means ± SEM are shown. n = 3, 3, 4 and 5 mice, respectively.

(C and D) Immunofluorescence (C) on primary hepatocytes isolated from SKO and Shp2/Prom1 double KO (DKO) mouse livers and quantification of Ki67+ ratio (D). Images of representative colonies are shown. ***P < 0.001, (two-tailed unpaired t test). Means ± SD from 4 wells are shown.

(E) Experimental design with primary hepatocytes isolated from GFP labeled SKO liver and unlabeled DKO liver. E-Cad+ colonies were analyzed.

(F) Immunofluorescence images of E-Cad+ colonies in SKO, DKO and mixed culture as shown in (E). The arrows show a GFP+ SKO cell forming a part of the colony with the surrounding DKO cells.

(G) Quantification of the Ki67 ratio in E-Cad+ colony-forming cells shown in (F). ***P < 0.1 (two-tailed unpaired t test). n.s., not statistically significant. Means ± SD from 3 wells are shown.

(H and I) Immunofluorescence (H) on WT and Prom1 KO mouse intestinal organoids treated with MEK inhibitor (MEKi) and quantification of Ki67+ ratio in the crypt cells (I). Dashed lines indicate the crypt buds. *P < 0.05, (two-tailed unpaired t test). Means ± SD from 3 wells are shown. Each dot represents each crypt buds, and symbols indicate each well.

Scale bars, 100 μm (A, C and F) and 50 μm (H).

Liver regeneration of WT and SKO livers after PHx

(A) General morphology of livers 2 days after PHx.

(B) H&E staining of liver sections 2 days after PHx.

(C and D) Ki67+ percentages in HNF4α+ hepatocytes (C) and HNF4α− NPCs (D) 2 days after PHx. Means ± SEM are shown, n=3 per group, **P < 0.01 (two-tailed unpaired t test). n.s., not significant.

(E) EpCAM was highly expressed in biliary epithelial cells (arrow), but not in the colony (dashed line). Stellate cells (GFAP) were not altered around the colony.

(F) CK19 was highly expressed in biliary epithelial cells (arrow), but not in the colony (dashed line). CD133 was highly expressed in the colony.

(G) Sox9 and CD44 were highly expressed in biliary epithelial cells (arrowhead), but was not upregulated in the colony (dashed line) compared to surrounding hepatocytes.

(H) AFP showed no difference between the colony (dashed line) and the surrounding tissue.

(I) Liver/body weight ratios after PHx. Means ± SEM from 3 or more mice analyzed for each time point and group are shown.

(J and K) Immunofluorescent staining of SKO liver sections 3 weeks after PHx. Colonies were at various sizes and locations as shown by arrows in (J). Macroscopic colonies were found as shown by arrows and dashed lines in (K). Immunofluorescent image in (K) corresponds to the magenta arrow in the liver image. PV, portal vein; CV, central vein.

(L) Vasculature shown by PECAM did not distinguish the colony (arrow).

(M) HGF expressed by NPCs was not concentrated in the colony (dashed line). Scale bars, 1 cm (A) 100 μm (B, E-G, I-L).

Clonal analysis of hepatocytes

(A) Predicted outcomes of different dynamics.

(B) Colony size/clone size ratios can be used to roughly estimate the original colony sizes.

(C) Immunofluorescence of SKO liver sections 2 days after PHx, showing mixture of binuclear and mononuclear hepatocytes in the colony.

(D) Immunofluorescence of primary hepatocyte cultures from WT and SKO mice. The arrow indicates an in vitro colony in Shp2-deficient hepatocyte culture. Arrowheads, dying cells with high autofluorescence.

Scale bars, 50 μm (C) and 100 μm (D).

Molecular analysis of intercellular signals during colony induction

(A) Schematic illustration of mosaic rescue experiment of Shp2 in Shp2-deficient liver.

(B) Immunofluorescence showing Shp2-rescued hepatocytes also labeled with GFP.

(C) GFP+ hepatocytes were proliferating (Ki67+; arrows), while other proliferating cells were not hepatocytes (arrowheads). Areas with colonies were not included here.

(D) Colonies were frequently constituted by GFP-negative hepatocytes (dashed line). Orange dots are non-specific stains of debris caused by HTVi.

(E) Some GFP+ hepatocytes showed expression of CD133 (arrowheads). Dashed lines show CD133-positive areas.

(F) qRT-PCR analysis of WT and SKO liver lysates 2 days after PHx. n=3, means ± SEM are shown, *P < 0.05, **P < 0.01 (two-tailed unpaired t test).

(G) Immunofluorescence on SKO liver section 2 days after PHx. White dashed lines mark Porcupine+ cells enriched in CD133+ colonies, rarely detected in non-colony area (arrowheads). Pink dashed lines: vasculature.

(H) Immunofluorescence on the colonies (arrows) in Shp2-deficient hepatocyte culture in vitro. Arrowhead, dying cells with high autofluorescence.

(I) Immunoblot analysis of WT and SKO liver lysates 2 days after PHx.

(J) qRT-PCR analysis of SKO liver tissue lysates without PHx (0d, 3 mice) and with PHx (2d, 3 mice). Means ± SEM are shown. **P < 0.01 (two-tailed unpaired t test).

(K) qRT-PCR analysis of hepatocyte and NPC fractions from SKO livers (3 mice) 2 days after PHx. HNF4α and CD45 were used as positive controls for the fractionation. β-actin was used for normalization, because GAPDH was highly expressed by hepatocytes. Means ± SEM are shown. *P < 0.05, **P < 0.01, ***P < 0.001 (two-tailed unpaired t test).

(L) qPCR analyses of PLC cells treated with Mek inhibitor and Porcn inhibitor. Means ± SD from 4 wells are shown. *P < 0.05, **P < 0.01, (two-tailed unpaired t test). n.s., not significant.

Scale bars, 100 μm (B-E, G and H).

Analyses of CD133 localization

(A-C) Immunofluorescence on PLC cells transfected with CD133-Myc-tag fusion protein expression construct. Cells with different expression levels are shown as representatives. In (A) and (C), exposure times for the CD133 signals were adjusted separately to clearly demonstrate the patterns, rather than intensities. Note that cells with higher CD133 expression displayed bulky pattern of the filaments, with maintained colocalization with tubulin filaments. Arrowheads, CD133+ filaments. As shown in (B), the overexpressed CD133 primarily localize to the filaments, without detectable membrane localization (arrowheads). With extreme overexpression, CD133 can also localize to the cellular membrane surface, altering the morphology of the surface (right panels).

(D) Immunofluorescence on HeLa and MC38 cells.

(E) Colocalization of CD133 and β-tubulin was analyzed as Pearson’s coefficient. Mismatched green and magenta channels from shuffled ROIs were measured as controls.

(F) Immunofluorescence on PLC cells transfected with cMet-GFP. cMet, an HGF receptor, shows the cell surface. CD133 did not colocalize with cMet-GFP on the cell surface, but instead localized to the filaments.

(G) Immuno-Gold EM images of cryo-ultramicrotome sections of SKO liver tissue after PHx. Light green arrowheads, CD133 staining (12nm colloidal gold); Magenta arrowheads, α-tubulin staining (18nm colloidal gold).

Scale bars, 5 μm (A and C), 50μm (B and D), 25μm (F) and 50nm (G).

Analyses of CD133+ vesicles

(A) Immunofluorescence for endosomal markers on PLC cells.

(B) Immunoblotting of CD133+ vesicles isolated from SKO liver after PHx. CHMP2B is a protein that is generally involved in vesicle formation.

(C) qRT-PCR analysis of housekeeping genes in RNAs extracted from WT (3 mice) and SKO tissues (3 mice) and from CD133+ vesicles and CD133-negative fractions (4 mice for both). Means ± SEM are shown.

(D) qRT-PCR analysis of RNAs extracted from CD133+ vesicles from PLC cells treated with MEK inhibitor.

(E) RNA-seq analysis showing enrichment scores of IEG contents in the two vesicle types compared to the whole cells. Each lane represents independent isolation.

(F) Immunofluorescence on SKO liver 2 days after PHx. Asterisks indicates vasculature.

(G) Immunofluorescence on MCF10A cells treated with inhibitors.

(H) qRT-PCR analysis of PLC cell lysates after treatment with CD133+ vesicles isolated from MEK inhibitor-treated PLC cells. RNase and Triton X-100 were used to digest the RNA content of the vesicles. *P < 0.05, **P < 0.01 (two-tailed unpaired t test). n.s., not statistically significant. Means ± SD from 3 replicates are shown.

Scale bars, 25μm (A) and 50μm (F and G).

Identification of hepatocytes in the scRNA-seq data set

UMAP analysis of the isolated cells from indicated livers. Hepatocyte markers and NPC markers were analyzed after the unsupervised clustering.

Single cell RNA-seq analysis of WT and SKO livers after PHx

(A) Heatmap analysis of single hepatocytes.

(B) Detection of CD133 expression in the single cell RNA-seq data.

(C) Expression levels of indicated hepatocyte subtype markers and stem cell-like markers. CD133- and CD133+ hepatocytes in the SKO liver 2 days after PHx were compared.

(D) Principal component analysis with the IEGs.

(E) tSNE analysis. Clusters likely reflect spatial locations within the tissue.

Simulation analysis of the IEG exchange with scRNA-seq of the regenerating SKO liver

(A) Possible models tested for the simulation. Examples with 5 communicating cells are shown. X is a variable that reflects the amount of RNA exchanged between each pair of cells. (B and C) Simulation with different RNA exchange model, different variable X and cell numbers.

CD133 is required for proliferation of crypt cells in the intestinal organoids during signal deficit

Immunofluorescence on WT and Prom1 KO mouse intestinal tissue sections and intestinal organoids. Scale bars, 20 μm.