Ductal heterogeneity in the adult zebrafish pancreas.

(A) Uniform Manifold Approximation and Projection (UMAP) plot showing the different cell clusters. The classifications were based on known marker genes highly expressed in each cell type in the zebrafish pancreas. The two clusters of pancreatic ductal cell types are denoted by dashed lines and arrows. NK, natural killer. (B and C) Feature plot shown in UMAP embedding of selected ductal markers (B), Notch and BMP downstream genes enriched in pancreatic ducts, and novel ductal markers (C). (D) Immunostaining using a Vasnb antibody in Tg(tp1:EGFP) transgenic larva showing the ducts in the hepatopancreatic biliary system. The dashed lines outline the organs in different colors, i.e., pancreas (yellow), gallbladder (magenta), liver (cyan), and intestine (white). Scale bars, 200 μm. (E to G) Representative confocal images of immunostaining of Cdh17 and Cldnc in Tg(tp1:EGFP) larvae (E) and (F) and in situ hybridization for anxa4 in Tg(tp1:EGFP) larvae (G) with Z-projection (E′) and (E″), (F′) and (F″), and (G′). The pancreas (yellow), islet (cyan), and non–Notch responsive ducts (white) were outlined by dashed lines in different colors. The Notch-responsive intrapancreatic ducts are indicated by arrows (orange). Scale bars, 200 and 40 μm (magnification). (H) Double transgenic for tp1:EGFP and tp1:H2BmCherry with anti-Vasnb staining showing the labeling pattern in the duct and islets in the larvae pancreas. The pancreata, extrapancreatic duct, intermediate duct, and islet are highlighted with yellow, cyan, white, and orange dashed lines, respectively. Scale bars, 200 μm (left) and 40 μm (right). A, anterior; P, posterior; D, dorsal; V, ventral. (I) In situ hybridization for krt4 showing the extrapancreatic-to-intermediate duct system with bifurcation in 6 dpf larva. The insets are magnified Z-projections with white dashed lines outlining the tree-like structure of the intermediate duct. Scale bars, 200 and 40 μm (insets).

Lineage tracing showed limited Notch-responsive duct–to–endocrine cells conversion at the principal islet.

(A and B) Tg(tp1:CreERT2) was crossed to two different color-switch lines under the control of ubiquitin B promoter. H2BmCherry/mCherry expression was specifically induced in tp1⁺ Notch-responsive ducts and their progenies with 20 μM 4-OHT administered from 3 to 4 dpf, as visualized with Z-projections 6 dpf. Yellow, white, and cyan dashed lines indicate the pancreata, Notch-responsive intrapancreatic ducts, and non–Notch-responsive duct anterior to the principal islet, respectively. Scale bars, 100 μm. (C and D) Schematics and confocal Z-projections of tp1 lineage–traced cells under β cell ablation condition. Tentative overlap in the Z-projection (the mCherry⁺ cell indicated by orange arrow) was confirmed to not coexpress Sst1.1 and insulin (Ins) in single planes. Scale bars, 40 μm. The yellow and blue dashed lines indicate the principal islet and extra- and intermediate pancreatic duct, respectively. (E and F) Long-term lineage tracing in tp1:CreERT2;ubi:Switch with 4-OHT treatment 3 to 4 dpf. Confocal Z-projections of pancreatic head (E) and tail (F) 23 dpf. Yellow dashed lines indicate the principal islet. Scale bars, 40 μm. (G and H) cftr:Gal4;UAS:EGFP;tp1:H2BmCherry indicate cftr⁺ cells 6 dpf (G) and 22 dpf (H). The ductal cells are highlighted using cyan dashed lines. Scale bars, 40 μm. (I and J) cftr:Gal4;UAS:EGFP;UAS:Cre;ubi:Switch indicate cftr⁺ lineage traced cells 14 dpf (I) and 20 dpf (J). Yellow dashed lines in (I) indicate the pancreatic head (magnifications in right panels). The yellow arrow points to lineage-traced endocrine cells. Yellow arrows in (J) indicate lineage-traced Gcg⁺ cells in head and tail regions. Scale bars, 40 μm.

Lineage tracing and targeted cell ablation using krt4 knock-in iCre line.

(A and B) Confocal images of krt4⁺ ductal cells in 6 dpf larvae (A) and 30 dpf juvenile fish (B). The brown, white, and cyan dashed lines indicate the pancreata, intestine, and the principal islet, respectively. The yellow arrows indicate large luminal krt4⁺ ducts throughout the pancreas, i.e., even in the pancreatic tail as the fish grow older. Scale bars, 100 μm. (C) The Cre/loxP strategy used to demonstrate krt4⁺ ductal cell–derived β cells. krt4-p2a-mNeonGreen-t2a-iCre was crossed to a color-switch line under the control of the insulin promoter. H2BmCherry is specifically induced in krt4-derived β cells, while β cells from other origins express CFP. (D and E) Z-projection showing differential distribution of β cells from krt4⁻ (green) and krt4⁺ (magenta) origins, 21 dpf (D) and 25 dpf (E). The ductal trees are visualized with anti-Vasnb staining. Scattered β cells budding out from the intermediate duct are highlighted with red arrows. White dashed lines indicate the principal islet. Scale bars, 40 μm. (F) Confocal images of krt4-derived β cells (magenta) in the secondary islets. Scale bars, 40 μm. (G to I) Design of targeted cell ablation of krt4-derived β cells by induction of DTA (G) with representative confocal images [(H) and (I)]. Cyan and yellow dashed lines in the Z-projection (H) indicate the area with surviving β cells at the posterior part of the principal islet and the β cell debris at the anterior part, respectively. The cell with orange arrow in (I) and magnified image (I′) is activated Caspase-3–positive, shown in single plane, indicating apoptosis. Scale bars, 80 μm.

Spatiotemporal-controlled lineage tracing of krt4⁺ ducts.

(A) Workflow of the Cre/loxP system used for spatiotemporal lineage tracing of krt4⁺ ductal cells. Fish carrying krt4-p2a-EGFP-t2a-CreERT2 (identified by green skin) are crossed to fish carrying ubi:CSHm transgenic lines (identified by blue skin). The 4-OHT is administered from 1 to 2 dpf, and we analyzed 35 dpf juvenile fish. (B) The maximum Z-projection of confocal images showing the islets from TgKI(krt4:CreERT2);Tg(ins:H2BGFP);Tg(ins:CSHm) line stained with Vasnb antibody (white) and 4′,6-diamidino-2-phenylindole (DAPI). All β cells are displayed with green fluorescence (from the ins:H2BGFP transgene) in the nucleus, and krt4-derived endocrine and ductal cells that can be traced back to krt4⁺ duct origin expressed H2BmCherry. Notably, the luminal duct in the pancreatic tail and Notch-responsive ductal cells are devoid of H2BmCherry. The fluorescence in krt4⁺ ducts is not visible because of its low intensity and the used confocal microscopy setting. (C to F) Single-plane confocal images showing the labeling pattern in three different layers/regions, with the cyan and yellow outlined regions (C) magnified in (D) to (F). Scale bars, 80 μm [(B) and (C)] and 40 μm [(D) to (F)].

Spatiotemporal-controlled lineage tracing of nkx6.1⁺ ducts.

(A) Single-plane confocal images showing the labeling pattern in TgKI(nkx6.1:CreERT2);Tg(ubi:CSHm) treated with 4-OHT from 1 to 2 dpf, displaying five distinct regions with high magnifications shown on the right and bottom, respectively. The extrapancreatic and intermediate ducts are devoid of labeling, while the luminal ducts in the pancreatic tail and Notch-responsive intrapancreatic ducts can be traced back to the nkx6.1⁺ cell origin. The ductal cells residing in between the intermediate duct and the luminal duct in the tail regions are devoid of labeling. The quantification results of the proportion of lineage traced ductal cells are displayed in (A′) (extra-intermediate duct region), (A″) (proximal luminal duct), and (A‴) (distal luminal duct). (B) Single-plane confocal images showing luminal duct and lineage-traced β cells in a secondary islet. The luminal ducts in the pancreatic tails are H2BmCherry⁺. The coexpression of H2BGFP and H2BmCherry indicate β cells within the secondary islet that can be traced back to the nkx6.1⁺ cell origin. The quantification results of ins:H2BGFP–single positive cells and ins:H2BGFP/H2BmCherry–double positive cells are shown in (B′), with the proportion of lineage traced ins:H2BGFP⁺ cell shown in (B″). Five size-matched secondary islets were used for the quantification. Anti-Vasnb (white) and ins:H2BGFP transgene (green) were used to visualize the ductal tree and the islets. Scale bars, 80 μm.

Single-cell transcriptomics highlight distinct molecular signatures in various cell types.

(A) t-distributed stochastic neighbor embedding (t-SNE) plot showing the cell type assignment of all single cells (~40,000). The classifications were based on previously known marker genes that were significantly enriched in each cluster (highlighted in parentheses). T_H2, T helper 2 cells; ILC2, innate lymphoid cells type 2. (B) t-SNE plots of well annotated cell markers, colored by the normalized gene expression levels. (C and D) UMAP plot of subclustering including pancreatic duct, krt5⁺ cells, and endocrine cells (~10,000). (E) UMAP plots of various marker genes, colored by the normalized gene expression levels. (F and G) UMAP plot and 3D PCA plot showing the embedding of krt5⁺ cell and all endocrine cells. (H) Heatmap highlighting the key lineage-committed transcription factors and hormones in each cluster, with colors displayed on column-scaled mean expression. The rows and clusters were ordered by hierarchical clustering of scaled expression values.

In silico analyses of transition-to-ins⁺/sst1.1⁺ hybrid cells.

(A) UMAP plot displaying three clusters of transition-to-ins⁺/sst1.1⁺ hybrid cells, color-coded by subpopulation. The numbers and proportions of each cell type were displayed in a pie chart. (B) The Slingshot and Monocle 3 analyses reveal a linear trajectory, with early transition cells predominantly early in pseudotime (left), late transition cells in the middle, and ins⁺/sst⁺ hybrid cells predominantly later in pseudotime (right). (C) Heatmap of the top 150 differentially expressed genes per cluster, with cell markers gene, DNA binding genes, tight junction genes, membrane protein gene, and mitochondrial genes highlighted on the right. (D) Gene Ontology (GO) analysis of differential expressed genes in each cell cluster along the transition-to-ins⁺/sst1.1⁺ hybrid cells. Dot plots showing the differentially enriched GO terms colored by adjusted P value and sized by gene ratio. (E) Ternary plot shows the expression of DNA binding genes in early transition, late transition, and ins⁺/sst1.1⁺ hybrid cells. Genes with highest percentage of expression in early transition cells are in blue, Late transition in green, and ins⁺/sst1.1⁺ hybrid cells in red. (F) Heatmaps showing the enriched regulons among the three cell states in the trajectory. The regulons were ordered by hierarchical clustering. (G) Heatmap showing the hierarchical clustering of identified regulon and four regulon modules. (H) The UMAP for transition-to-ins⁺/sst⁺ hybrid cells based on the regulon activity scores (RASs), each cell is color-coded on the basis of the cell states assignment. (I) UMAP projection of transition-to-ins⁺/sst1.1⁺ hybrid cells overlayed by the expression level of transcription factors (left) and RAS (right).

Velocity-based analyses and pathway validation.

(A) UMAP colored by latent time progression. Cells later in the trajectory (determined by RNA velocity) are yellow. The UMAP is overlaid by arrows indicating the extrapolated future cell states identified through RNA velocity analysis, suggesting that both differentiation and dedifferentiation processes occur simultaneously during insulin-producing cell recovery. (B) UMAP projection of transition-to-ins⁺/sst1.1⁺ hybrid cells with eight subclusters (three early transition cell states, two late transition cell states, and three ins⁺/sst1.1⁺ hybrid cell states). (C) The PAGA analysis indicates dedifferentiation process. (D and E) The root identification and velocity-based pseudotime results by CellRank. (F) Graphical summary using Sankey plots demonstrating the relationship between cell states (left) and condition (right). (G) Representative single-plane confocal image of hes6⁺ cells from the immunostaining of the knock-in hes6-p2a-EGFP-t2a-CreERT zebrafish line. The yellow arrows point to two EGFP⁺ cells adjacent to the duct. (H) Scatterplots showing the level of unspliced versus spliced transcripts of cell state specific genes. (I) GO analysis of up-regulated genes expressed in late transition 1 and late transition 2 cells, with enriched gene shown below. Dot plots showing the differentially enriched GO terms colored by adjusted P value and sized by gene ratio. MAPK, mitogen-activated protein kinase. (J) Violin plots and dot plots highlighting the differential gene set scores between late transition 1 and late transition 2 cells using single-cell gene set variation analysis. (K) UMAP plot showing the statistically significant gene activity scores in each cell. IL-6, interleukin-6; Jak, Janus kinase; Stat3, signal transducers and activators of transcription 3; AUC, area under the curve. (L) Representative single-plane confocal images and quantifications of β cells following treatment with either dimethyl sulfoxide (DMSO) or the PI3K inhibitor wortmannin. Treatment with wortmannin increased the number of newly generated krt4-derived β cells (mCherry⁺/insulin⁺; Wilcoxon test) in the basal state (without β cell ablation). Scale bars, 40 μm.

Graphical summary describes the landscape of endocrinogenesis in zebrafish pancreas.

(A) Illustration of zebrafish pancreas in larval and juvenile/adult stage. The schematic showing that in larval stage, the zebrafish ductal tree is composed of extrapancreatic duct (EPD) (foxj1a^+/krt4⁺), intermediate duct (foxj1^−/krt4⁺), and Notch-responsive intrapancreatic duct (IPD) (tp1^+/krt4⁻); while in juvenile and adult, the krt4⁺ duct forms luminal structure extending to the pancreatic tail, which also contain Notch-responsive ductal cells located in the peripheral regions. (B) The sketch of zebrafish endocrinogenesis suggesting multiple phases of neogenesis that take place at various stages. (C) Schematic summary of the divergent origins and contributions of the krt4⁺duct in juvenile and adult pancreata. (D and E) The working model of β cell differentiation and insulin-producing cell regeneration. During development, the primitive islet is formed from the dorsal pancreatic bud, followed by krt4-derived duct-to-endocrine differentiation in the principal islet. The Notch-responsive duct does not naturally serve as the direct progenitor of endocrine cells but gradually undergoes transformation and assembles to become krt4⁺ luminal duct, which has neogenic competence and can differentiate into endocrine cells in the secondary islets throughout the development until the adult stage. After massive loss of β cells, the neurod1⁺/sst1.1⁺ endocrine cells would initiate reprogramming and up-regulate insulin expression for functional compensation.