Fig 2: A+ ß-cells are converted to A– ß-cells.A Generation of Aldh1a3-Creert:YFPfl/+ db/db mice and tracing of A+ cells in db/db mice during pair-feeding. B Schematic drawing of the experimental procedure. C IPGTT after 4 weeks of pair-feeding (n = 6, 5, 4 or 10 for ad lib, PF_R (Pair-fed responder), PF_NR (Pair-fed non-responder) or lean mice. Data are expressed as means ± SEM. Two-way ANOVA with multiple comparison test was used for statistical analysis. The P value was 0.0097, 0.0077, 0.0002 for time points 0, 30, or 60 min and P < 0.0001 for 90 and 120 min in comparison of ad lib with PF_R. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. D–F FACS to detect lineage-traced A+ cells before Pair-feeding (YFP+) and cells with ongoing ALDH1A3 activity (Red+) in Ad lib (D), PF_NR (E) or PF_R (F). G, H Quantification of YFP+ (G) cells or YFP+/ongoing ALDH1A3+ cells (H). All data are expressed as means ± SEM for n = 5, 4, or 5 biologically independent samples per group. One-way ANOVA with multiple comparison test was used for statistical analysis. I Immunofluorescent staining for ALDH1A3, YFP, PDX1 and Insulin using isolated islets in each condition as in (A). Representative immunofluorescence images of n = 4 mice per group. Scale bars: 20 µM. Source data are provided as a Source Data file.

Fig 3: ß-Aldh1a3 KO improves diabetic phenotypes.A Glucose levels in fasted and refed 12-week-old male mice (n = 5 mice per genotype). B IPGTT in 20-week-old mice (n = 5 per genotype). The P value was 1.6E–8, 1.0E–15, 2.6E–7, 3.3E–10, 1.0E–15, or 1.0E–15, for each time point in comparison of Aldh1a3 KO_db/db with db/db mice. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. C Plasma insulin levels in fasted and refed 12-week-old male mice (n = 5 per genotype). D Insulin secretion in islets from ß-Aldh1a3 KO_db/db, lean or db/db mice. ANOVA was performed between the two groups (n = 8 per group). E Immunofluorescent staining of PDX1, INS, E-Cadherin and ALDH1A3 in ß-Aldh1a3 KO_db/db or db/db mice. Representative immunofluorescence images of n = 3 mice per group. Scale bars: 50 µM. All data are expressed as means ± SEM. Two-way ANOVA with multiple comparison test was used for statistical analysis for (A–D). Source data are provided as a Source Data file.

Fig 4: A. List of top candidate compounds from immunofluorescence screen with their respective fold changes for insulin and glucagon expression, and for the number of DAPI-positive nuclei, as measured via immunofluorescence intensity. B. Loperamide dose ressponse suppression of glucagon transcription in Min6 cells treated with FoxO inhibitor. C. Glucagon mRNA levels in Min6 cells following loperamide time course, as measured by RT-qPCR. D. PPAR-gamma and Aldh1a3 transcription in Min6 cells treated with FoxOi/loperamide for 48h, measured by RNAseq. E. Gene ontology enrichment terms for RNAseq dataset of genes significantly downregulated by FoxOi and significantly rescued by loperamide after 48h treatment in Min6 cells.

Fig 5: Model of ALDH expression and retinol association with FRDA cardiomyopathy. Expression of several enzymes necessary for biosynthesis of RA is altered in FRDA fibroblasts. ALDH1A3 expression is higher in FRDA samples than in age- and sex-matched controls and is especially increased in FRDA samples derived from individuals without a CMP diagnosis. ALDH, aldehyde dehydrogenase; ALDH1A3, aldehyde dehydrogenase family 1 member A3; CMP, cardiomyopathy; FRDA, Friedreich's ataxia; RA, retinoic acid.