Genetic screening in zebrafish identifies igfbp1a as a promoter of βcell regeneration

A. Schema of the analysis of gene expression in islets isolated from control larvae and larvae subjected to βcell ablation. β cells were ablated by exposing nitroreductase (NTR)expressing transgenic larvae to metronidazole (MTZ) from 3 to 4 dpf. Islets were then isolated, and their RNA extracted and analyzed by microarray. Out of the 470 genes that were upregulated more than 50%, 33 genes encoded proteins that harbored a signal peptide for secretion (according to the algorithm of SignalP). Excluding genes that encode enzymes or proteins with a transmembrane (TM) domain, we selected 11 genes for overexpression studies in zebrafish larvae (C-E).

D, E. Representative images at 6 dpf of Tg(ins:kaede);Tg(ins:CFPNTR) transgenic larvae that had been injected at the 1-2 cell stage with transposase mRNA (control) or transposase mRNA + bactin:igfbp1a (bactin:igfbp1a), subjected to βcell ablation by metronidazole (MTZ) during 3-4 dpf, and subsequently allowed to regenerate for 2 days. The GFP⁺ heart (arrowhead) visualizes successful integration of the construct. Islets are indicated by white arrows. Scale bars: 100 µm.

F. Quantification of βcell regeneration at 6 dpf in control (n = 23), bactin:igfbp1aoverexpressing (n = 13), and bactin:igfbp1boverexpressing (n = 8) Tg(ins:kaede);Tg(ins:CFPNTR) larvae; ***P = 0.0002, ns = nonsignificant (P = 0.3106).

H. Both liverspecific (lfabp promoter; n = 46) and widespread (bactin promoter; n = 18) overexpression of igfbp1a increase βcell regeneration when compared to control (n = 77); ***P < 0.001.

I-M. Quantification of β cells with or without βcell ablation and igfbp1a overexpression by confocal microscopy, which detects even weakly insulinexpressing β cells; ****P < 0.0001. Scale bars: 15 µm.

Igfbp1a′s effect on β-cell regeneration is specific and functionally relevant

A-C. Igfbp1a promotes βcell regeneration, rather than βcell survival. To trace β cells, we exposed Tg(ins:kaede) larvae to UV light (which causes the existing Kaede protein to switch from emitting green fluorescence to emitting red fluorescence) just before ablating the β cells by MTZ treatment from 3 to 4 dpf. After regeneration, the newly formed β cells are green, whereas the β cells that survived ablation are yellow (overlap of green and red). Representative confocal images (A, B) at 6 dpf of control and bactin:igfbp1a-overexpressing Tg(ins:kaede);Tg(ins:CFPNTR) transgenic larvae; arrows indicate surviving (yellow) β cells. Scale bars: 10 µm. (C) Quantification of βcell regeneration (green bars) and βcell survival (yellow bars) per larva at 6 dpf. ***P < 0.001; n = 20 larvae in the control group, n = 27 larvae in the bactin:igfbp1a-overexpressing group.

D-F. Igfbp1a does not promote δcell regeneration. We treated control and bactin:igfbp1a-overexpressing Tg(sst:flagNTR);Tg(sst:dsRed) larvae with MTZ from 3 to 4 dpf to ablate δ cells, and then allowed them to regenerate until 6 dpf. Representative confocal images (D, E) at 6 dpf of control and bactin:igfbp1a-overexpressing larvae showing comparable number of δ cells after 2 days of regeneration. Scale bars: 15 µm. (F) Quantification of the total number of δ cells per δcellablated larva at 6 dpf compared to the baseline number of δ cells in nonablated control larvae. ns, P = 0.2325; n = 13 in the control group, n = 7 in the bactin:igfbp1a group.

G. Other Igfbps do not promote βcell regeneration. We injected Tg(ins:kaede);Tg(ins:CFPNTR) transgenics with either transposase mRNA (control; n = 31) or transposase mRNA + one of six different igfbps, igfbp2a (n = 25), igfbp2b (n = 25), igfbp3 (n = 26), igfbp5a (n = 17), igfbp6b (n = 34), and igfbp7 (n = 17), treated them with MTZ from 3 to 4 dpf to ablate their β cells, and then quantified their β cells after 2 days of regeneration, at 6 dpf; ns=nonsignificant (igfbp2a P = 0.986, igfbp3 P = 0.979, igfbp5a P = 0.999, igfbp6b P = 0.997, and igfbp7 P = 0.999); igfbp2b *P = 0.037.

H. Tg(pcsk1:GFP) is expressed in regenerating β cells within the islet (arrowheads), but not outside the islet (arrow), at 6 dpf in larvae overexpressing igfbp1a. Scale bar: 15 µm.

I, J. Freeglucose levels during βcell regeneration in control (bactin:mCherry) and bactin:igfbp1a-overexpressing Tg(ins:kaede);Tg(ins:CFPNTR) larvae (I), as well as in Tg(ins:kaede);Tg(ins:CFPNTR) larvae injected in the pericardial sac at 4 dpf with 1 ng of recombinant mouse Igfbp1 (J). We treated larvae with MTZ from 3 to 4 dpf to ablate their β cells and monitored their freeglucose levels at 3-7 dpf. Freeglucose levels were significantly lower after genetic igfbp1a overexpression or Igfbp1protein injection (red lines) than in controls (black lines) at 7 and 6 dpf, respectively. Baseline reference levels of free glucose throughout development are shown for a different set of larvae without βcell ablation. n = 24 larvae (four pools of six larvae) per data point; ***P < 0.001, *P < 0.05.

Cellular mechanisms of Igfbp1a′s effect on βcell regeneration

A-C. Tg(ins:GFP);Tg(tp1:H2BmCherry);Tg(ins:FlagNTR) transgenics, with or without Tg(bactin:igfbp1a), were treated with MTZ from 3 to 4 dpf to ablate the β cells and were then allowed to regenerate from 4 to 6 dpf. Representative confocal images at 6 dpf of control (A) and Tg(bactin:igfbp1a) (B) larvae showing a modest number of ins⁺ tp1⁺ coexpressing cells, indicated by arrows, after 2 days of regeneration. Scale bars: 15 µm. (C) Quantification of the total number of β cells (green bars) at 6 dpf, and of β cells expressing Tg(tp1:H2BmCherry), that is, of ductal origin (yellow bars). ****P < 0.0001, ns= nonsignificant. n = 23 larvae in the control group, n = 17 larvae in the Tg(bactin:igfbp1a) group.

D-F. Tg(ins:VenuszGeminin) was examined in control and Tg(bactin:igfbp1a) larvae at 6 dpf, after βcell ablation 3-4 dpf by using Tg(ins:FlagNTR). Representative confocal images (D, E) at 6 dpf of control and Tg(bactin:igfbp1a) larvae showing cell cycle activation of β cells in green. Scale bars: 15 µm. (F) Quantification of the total number of β cells with activated cell cycle at 6 dpf, ***P < 0.001, n = 32 larvae in the control group, n = 41 larvae in the Tg(bactin:igfbp1a) group. The number of β cells with activated cell cycle is displayed together with the total number of β cells in experiments with the same setup, ****P < 0.0001. n = 39 larvae in the control group, n = 33 larvae in the Tg(bactin:igfbp1a) group.

J-L. Tg(ins:kaede);Tg(ins:CFPNTR) transgenics, with or without Tg(bactin:igfbp1a), were injected with a control morpholino or a morpholino that knocked down arx, which is necessary for αcell differentiation and thus glucagon expression (J, K). Scale bars: 20 µm. (L) Quantification of the total number of regenerating β cells at 4 dpf (after βcell ablation at 2-3 dpf). ****P < 0.0001, *P < 0.05, ns=nonsignificant. n = 42, 20, 48 and 20, respectively.

Igfbp1a increases βcell regeneration by potentiating α to βcell transdifferentiation

A-E. Control and Tg(bactin:igfbp1a)overexpressing larvae were treated with MTZ from 3 to 4 dpf to ablate β cells and analyzed at 6 dpf, after 2 days of regeneration. Representative confocal images (A, B) of control and Tg(bactin:igfbp1a)overexpressing Tg(ins:dsRed);Tg(gcg:GFP);Tg(ins:FlagNTR) larvae. A bihormonal glucagon and insulinexpressing cell (gcg⁺;ins⁺) is indicated by an arrow. Representative confocal images (C, D) of control and Tg(bactin:igfbp1a)overexpressing Tg(gcg:GFP);Tg(ins:FlagNTR) larvae stained for Pdx1. Pdx1 and glucagonexpressing cells (pdx1⁺;gcg⁺) are indicated by arrows. Scale bars: 20 µm. (E) Quantification of gcg⁺;ins⁺ and pdx1⁺;gcg⁺ cells in the control and Tg(bactin:igfbp1a) group; ***P < 0.001, **P = 0.0012; n = 18 and 18 in the control groups, n = 15 and 10 in the Tg(bactin:igfbp1a) groups, for the gcg⁺;ins⁺ and pdx1⁺;gcg⁺ quantification, respectively.

F-I. Lineagetracing evidence supports α to βcell transdifferentiation. (F) Schema for temporal and spatial lineage tracing. loxPmediated excision of the STOP cassette permits heatinducible expression of the stable fusionprotein H2BGFP. Control and bactin:igfbp1aoverexpressing Tg(hsp:loxPmCherrySTOPloxPH2BGFP);Tg(gcg:CRE);Tg(ins:FlagNTR) larvae were first heatshocked at 3 dpf to label the glucagonexpressing cells with the stable fluorescent protein H2BGFP, thereafter treated with MTZ from 3 to 4 dpf to ablate the β cells, and analyzed at 6 dpf after 2 days of regeneration. Note that the temporal control of the lineage tracing is mediated by a heatshock. (G) Quantification of insulinexpressing cells that originate from α cells. Control larvae, n = 13; bactin:igfbp1a larvae, n = 16. **P = 0.0051. Representative confocal images (H, I) at 6 dpf (after 2 days of regeneration) of control and bactin:igfbp1aoverexpressing larvae stained for insulin, glucagon, topro, and H2BGFP. ins⁺;H2BGFP⁺ cells are indicated by the arrowheads, and H′ and I′ show staining against only insulin and H2BGFP. Moreover, most ins⁺;H2BGFP⁺ cells no longer express glucagon. Scale bars: 15 µm.

Inhibition of the Igf pathway mimics Igfbp1a′s stimulatory effect on βcell regeneration

A-H. PPP, an IGF1R inhibitor, promotes βcell regeneration. Tg(ins:H2BGFP);Tg(ins:FlagNTR) larvae were treated with MTZ from 3 to 4 dpf to ablate the β cells and then treated with EdU and DMSO or with EdU and PPP during regeneration from 4 to 6 dpf. Representative confocal images (A, B) at 6 dpf of DMSO and PPPtreated larvae displaying β cells in green and the β cells that had incorporated EdU as yellow overlap (arrowheads). (C) Quantification of the total number of β cells (green bars) at 6 dpf, and β cells that had incorporated EdU (white bars) from 4 to 6 dpf during βcell regeneration. ***P = 0.0003, P = 0.8607, respectively. (D) Rate of βcell proliferation displayed as the percentage of β cells that incorporated EdU. P = 0.1194. n = 18 larvae for the DMSOtreated group, n = 17 larvae for the PPPtreated group. (E-H) To examine whether PPP affected βcell proliferation during regular development, Tg(ins:H2BGFP) larvae were treated with EdU and DMSO or PPP from 4 to 6 dpf. Representative confocal images (E, F) of 6 dpf DMSO and PPPtreated larvae displaying β cells in green and the β cells that had incorporated EdU as yellow overlap. Scale bars: 20 µm. (G) Quantification of the total number of β cells (green bars) and β cells that had incorporated EdU (white bars) per larva from 4 to 6 dpf. P = 0.9098 and 0.9976, respectively. (H) Rate of βcell proliferation displayed as the percentage of β cells that incorporated EdU. P = 0.7822. n = 16 larvae for DMSOtreated group, 18 larvae for PPPtreated group.

I-K. Activation of the Igf pathway reduces βcell regeneration. Control and bactin:igf2aoverexpressing Tg(ins:H2BGFP);Tg(ins:FlagNTR) larvae were treated with MTZ from 3 to 4 dpf to ablate β cells and subsequently let to regenerate from 4 to 6 dpf. Representative confocal images (I, J) of 6 dpf control and bactin:igf2aoverexpressing larva displaying β cells in green. Scale bars: 15 µm. (K) Quantification of the total number of β cells per larva at 6 dpf following βcell regeneration from 4 to 6 dpf. ***P < 0.0001. n = 28 larvae for control, n = 15 larvae for bactin:igf2a.

L-P. No synergistic effect was observed for igfbp1a and PPP. Control and bactin:igfbp1aoverexpressing Tg(ins:H2BGFP);Tg(ins:FlagNTR) larvae were treated with MTZ from 3 to 4 dpf to ablate β cells and subsequently treated with DMSO or PPP during regeneration from 4 to 6 dpf. (L) Quantification of the total number of β cells per larva. n > 10 (n = 23, 20, 14, 13). P = 0.9546. Representative confocal images (M-P) of 6 dpf control or bactin:igfbp1a overexpressing larvae treated with either DMSO or PPP, displaying β cells after 2 days regeneration. Scale bars: 10 µm.

Igfbp1a also increases βcell regeneration in 1 month old zebrafish

A, B. Representative confocal projections of the whole pancreas (dashed lines) of 35dayold Tg(ins:H2BGFP);Tg(ins:FlagNTR) transgenics, with or without Tg(bactin:igfbp1a), that were subjected to βcell ablation between day 30 and 31 and then allowed to regenerate for 4 days. Scale bars represent 100 µm.

C. Body length of control and Tg(bactin:igfbp1a) zebrafish; ****P < 0.0001.

D-H. Quantification of βcell regeneration was automated with an ImageJ script. (D) Insulinpositive area per zebrafish. (E) Relative insulinpositive area per body length; **P < 0.01. (F) The percentage of the pancreas area that was insulinpositive was significantly larger in Tg(bactin:igfbp1a) than in controls; ****P < 0.0001. (G) The total number of recorded units of adjacent ins:H2BGFP⁺ pixels (units ranging from single β cells to βcell clusters) did not differ between control and Tg(bactin:igfbp1a) zebrafish. (H) The size of recorded ins:H2BGFP⁺ units was on average larger in Tg(bactin:igfbp1a) than in controls; *P = 0.039. n = 26 in the control group, n = 27 in the Tg(bactin:igfbp1a) group.

The IGF1R inhibitor PPP increases βcell regeneration and coexpression of insulin and pdx1 with glucagon

A-C. Pdx1 expression in 6dpf larvae during regeneration. Tg(ins:dsRed);Tg(gcg:GFP);Tg(ins:FlagNTR) larvae were treated with MTZ from 3 to 4 dpf to ablate the ² cells and subsequently treated with DMSO or the IGF1R inhibitor PPP from 4 to 6 dpf. Representative confocal images (A, B) at 6 dpf of DMSO or PPPtreated larvae. A gcg⁺ins⁺pdx1⁺ cell is indicated by an arrow. Scale bars: 15 µm. (C) Quantification of gcg⁺, gcg⁺ins⁺, and pdx1⁺gcg⁺ cells (P = 0.8409, *P = 0.0232, and ***P = 0.0007, respectively) in DMSO and PPPtreated groups. n = 20 for each group. Results are presented as mean values ± SEM and analyzed with ANOVA. Related to Fig 4.

The IGF1R inhibitor JB1 promotes βcell regeneration

A-D. Tg(ins:H2BGFP);Tg(ins:FlagNTR) larvae were treated with MTZ from 3 to 4 dpf to ablate the β cells. DMSO or JB1 was then injected into the larval pericardial cavity at 4 dpf, and the larvae were treated with EdU from 4 to 6 dpf. Representative confocal images (A, B) at 6 dpf of DMSO and JB1injected larvae, showing β cells in green and the β cells that incorporated EdU as yellow (green and red overlap; see arrowheads). Scale bars: 20 µm. (C) Quantification of the number of all β cells (green bars) and of β cells that incorporated EdU (white bars) per larva at 6 dpf. *P = 0.0422; ns=nonsignificant (P = 0.9944). (D) Rate of βcell proliferation, shown as the percentage of β cells that incorporated EdU. ns=nonsignificant (P = 0.5885). n = 12 larvae in the DMSOinjected group; n = 19 larvae in the JB1injected group.