LSD1-mediated enhancer silencing attenuates retinoic acid signalling during pancreatic endocrine cell development

Abstract

Developmental progression depends on temporally defined changes in gene expression mediated by transient exposure of lineage intermediates to signals in the progenitor niche. To determine whether cell-intrinsic epigenetic mechanisms contribute to signal-induced transcriptional responses, here we manipulate the signalling environment and activity of the histone demethylase LSD1 during differentiation of hESC-gut tube intermediates into pancreatic endocrine cells. We identify a transient requirement for LSD1 in endocrine cell differentiation spanning a short time-window early in pancreas development, a phenotype we reproduced in mice. Examination of enhancer and transcriptome landscapes revealed that LSD1 silences transiently active retinoic acid (RA)-induced enhancers and their target genes. Furthermore, prolonged RA exposure phenocopies LSD1 inhibition, suggesting that LSD1 regulates endocrine cell differentiation by limiting the duration of RA signalling. Our findings identify LSD1-mediated enhancer silencing as a cell-intrinsic epigenetic feedback mechanism by which the duration of the transcriptional response to a developmental signal is limited.

Fig. 1. Endocrine cell formation requires LSD1 activity during a short window in early pancreatic development.

a Schematic of the human embryonic stem cell (hESC) differentiation protocol to the endocrine cell stage (EN) and experimental plan for LSD1 inhibition. b Immunofluorescent staining for pancreatic hormones insulin (INS), glucagon (GCG) and somatostatin (SST) or PDX1 and NKX6.1 in control EN cells compared to EN cells with early (LSD1i^early) and late (LSD1i^late) LSD1 inhibition (representative images, n = 10 independent differentiations). Scale bar, 50 µm. c qRT-PCR analysis for INS, GCG and SST in control, LSD1i^early and LSD1i^late EN cells. Data are shown as mean ± S.E.M. (n = 3 replicates from independent differentiations with n = 3 technical replicates per sample; source data are provided as a Source Data file). P = 7.93 e−4, 1.42 e−2, 2.32 e−4, 8.71 e−4, 3.5 e−2, and 1.52 e−3, respectively, Student’s t-test, 2 sided. d Flow cytometry analysis at EN stage for NKX6.1, PDX1 and INS comparing control, LSD1i^early and LSD1i^late cells. Isotype control for each antibody is shown in red and target protein staining in green. Percentage of cells expressing each protein is indicated (representative experiment, n = 2 independent differentiations). D, day; AA, activin A; ITS, insulin-transferrin-selenium; TGFBi, TGFβ R1 kinase inhibitor; KC, KAAD-cyclopamine; KGF, keratinocyte growth factor; RA, retinoic acid; EGF, epidermal growth factor; ES, human embryonic stem cells; DE, definitive endoderm; GT, primitive gut tube; PP1, early pancreatic progenitors; PP2, late pancreatic progenitors; EN, endocrine cell stage; FSC-A, forward scatter area.

Fig. 2. LSD1 inhibition prevents decommissioning of transiently active early pancreatic enhancers.

a Heatmap showing density of ChIP-seq reads for LSD1 and H3K27ac centred on LSD1 peaks, spanning 10 kb. G1, G2 and G3 groups of LSD1-bound enhancers are deactivated (G1; n = 1345), remain active (G2; n = 765), or are activated (G3; n = 511) from PP1 to PP2. b Box plots of H3K4me1 and H3K4me2 ChIP-seq counts at G1, G2 and G3 enhancers at PP1 and PP2 stages. Plots are centred on median, with box encompassing 25th–75th percentile and whiskers extending up to 1.5 interquartile range (Tukey style). P = 5.0 e−12, < 2.2 e−16, <2.2 e−16, 1.73 e−2, <2.2 e−16, and 8.23 e−14, respectively, Wilcoxon rank-sum test, 2 sided. c Tag density plots displaying LSD1 tag distribution at G1, G2 and G3 enhancers at GT, PP1 and PP2 stages, centred on PP1 LSD1 peaks. d Tag density plots (left) for G1 enhancers displaying H3K27ac, H3K4me2 and H3K4me1 tag distribution at GT and PP1 stages, and at PP2 stage with and without early LSD1 inhibition (TCP, LSD1i^early). Plots are centred on PP1 LSD1 peaks. Box plots (right) of H3K27ac, H3K4me2 and H3K4me1 ChIP-seq counts at G1 enhancers at PP2 stage with and without early LSD1 inhibition (LSD1i^early). Plots are centred on median, with box encompassing 25th–75th percentile and whiskers extending up to 1.5 interquartile range (Tukey style). P = 4.59 e−5, <2.2 e−16, and <2.2 e−16, respectively, Wilcoxon rank-sum test, 2 sided. e LSD1, H3K4me2, and H3K27ac ChIP-seq profiles at an enhancer near HOXA1. f Model for LSD1-dependent enhancer decommissioning. Enhancer deactivation by removal of acetylation from H3K27 occurs independent of LSD1 activity. LSD1 subsequently mediates enhancer decommissioning by removal of H3K4me2 marks. KC, KAAD-cyclopamine; KGF, keratinocyte growth factor; RA, retinoic acid; EGF, epidermal growth factor. GT, primitive gut tube; PP1, early pancreatic progenitors; PP2, late pancreatic progenitors. All ChIP-seq experiments, n = 2 replicates from independent differentiations.

Fig. 3. LSD1 activity is necessary for repressing transiently expressed retinoic acid-dependent genes.

a Volcano plot of differentially expressed genes at PP2 after LSD1 inhibition from PP1 to PP2 (TCP, LSD1i^early). Differential expression calculated with DESeq2 and genes with ≥1.5-fold change up or down. Adjusted p-values of <0.05 were considered differentially expressed. 445 genes were downregulated and 955 were upregulated in LSD1i^early PP2 cells. Black dots indicate genes not significantly changed (p-value > 0.05), grey dots genes significantly changed (p-value < 0.05) but less than 1.5-fold compared to control, red and green dots genes significantly up- and downregulated (p-value < 0.05 and ≥1.5-fold change), respectively (n = 2 replicates from independent differentiations). b Enrichment analysis of genes upregulated by LSD1i^early within 100 kb of G1, G2 and G3 or other distal LSD1 peaks. Dashed lines indicate p-value = 0.05 for enrichment (positive value) or depletion (negative value), permutation test. c Percentage of LSD1i^early upregulated genes near G1 (n = 139), G2 (n = 78) and G3 (n = 53) enhancers (within 100 kb). d Box plot of mRNA levels for 139 LSD1i^early upregulated genes near G1 enhancers. Plots are centred on median, with box encompassing 25th–75th percentile and whiskers extending up to 1.5 interquartile range (Tukey style). P = 2.30 e−3, 4.38 e−6, and 2.25 e−7, respectively, Wilcoxon rank-sum test, 2 sided. e Gene ontology analysis for 139 LSD1i^early upregulated genes near G1 enhancers. GT, primitive gut tube; PP1, early pancreatic progenitors; PP2, late pancreatic progenitors.

Fig. 4. Prolonged retinoic acid exposure of early pancreatic progenitor cells phenocopies LSD1 inhibition.

a Enriched transcription factor (TF) binding motifs with associated p-values for G1 enhancers compared to G2 and G3 enhancers. Fisher’s exact test, 2 sided, corrected for multiple comparisons. b Enrichment for RXR peaks (±1 kb) among G1 enhancers versus random genomic regions. P = 0, permutation test. c Experimental plan to extend retinoic acid (RA) exposure through PP1 to PP2 (RA^extended) during hESC differentiation to the endocrine cell stage (EN). d Immunofluorescent staining for insulin (INS), glucagon (GCG) and somatostatin (SST) in control EN cells compared to EN cells with extended RA treatment (RA^extended) (representative images, n = 3 independent differentiations). Scale bar, 50 µm. e Flow cytometry analysis at EN stage for NKX6.1, PDX1 and INS comparing control and RA^extended cultures. Isotype control for each antibody is shown in red and target protein staining in green. Percentage of cells expressing each protein is indicated (representative experiment, n = 2 independent differentiations). f Volcano plot of differentially expressed genes at PP2 in RA^extended cultures. Differential expression calculated with DESeq2 and genes with ≥1.5-fold change up or down. Adjusted p-values < 0.05 were considered differentially expressed. 96 genes were downregulated and 69 were upregulated in RA^extended cultures. Black dots indicate genes not significantly changed (p-value > 0.05), grey dots genes significantly changed (p-value < 0.05) but less than 1.5-fold compared to control, red and green dots genes significantly up- and downregulated (p-value < 0.05 and ≥ 1.5-fold change), respectively. Yellow dots highlight genes also upregulated after LSD1 inhibition from PP1 to PP2 (TCP, LSD1i^early) (n = 2 replicates from independent differentiations). g Enrichment analysis of genes associated with LSD1-bound enhancers and upregulated by LSD1i^early among those upregulated by RA^extended. Dashed line indicates p-value = 0.05, Fisher’s exact test, 2 sided. h mRNA levels of select genes significantly upregulated in both LSD1i^early and RA^extended PP2 cells. Levels at PP1 stage are also displayed. Data shown as mean FPKM ± S.E.M. (n = 2 replicates from independent differentiations; source data are provided as a Source Data file). GT, primitive gut tube; PP1, early pancreatic progenitors; PP2, late pancreatic progenitors.

Fig. 5. LSD1 decreases future inducibility of RA-dependent genes by retinoic acid.

a Experimental plan to re-introduce retinoic acid (RA) during the PP2 to endocrine (EN) transition of hESC differentiation without (RA^late) and with prior inhibition of LSD1 (LSD1i^early + RA^late). The cartoon depicts the chromatin state as determined in Fig. 2. b Immunofluorescent staining for insulin (INS), glucagon (GCG) and somatostatin (SST) in control EN cells compared to EN cells with late RA treatment (RA^late) (representative images, n = 3 independent differentiations). Scale bar, 50 µm. c Flow cytometry analysis at EN stage for NKX6.1, PDX1 and INS comparing control and RA^late cells. Isotype control for each antibody is shown in red and target protein staining in green. Percentage of cells expressing each protein is indicated (representative experiment, n = 2 independent differentiations). d Gene expression changes in RA^late vs control EN cells compared to gene expression changes in LSD1i^early + RA^late vs LSD1i^early EN cells (calculated with DESeq2). Green dots indicate genes more increased in RA^late and red dots indicate genes more increased in LSD1i^early + RA^late compared to respective controls. Yellow dots highlight genes upregulated after both LSD1i^early and extended RA treatment (RA^extended) from PP1 to PP2. (n = 2 replicates from independent differentiations per condition). PP1, early pancreatic progenitors; PP2, late pancreatic progenitors; FSC-A, forward scatter area.

Fig. 6. Lsd1 is required for endocrine cell formation in mice during a short window in early pancreatic development.

a Strategy for conditional Lsd1 deletion in embryonic pancreatic progenitors of mice (Lsd1^Δpan mice). Yellow boxes: exons; green triangles: loxP sites. b Immunofluorescent staining for Pdx1 at embryonic day (e) 12.5, Lsd1 and chromogranin A (ChgA) or Sox9 and Ngn3 at e15.5, and Lsd1, insulin (Ins) and glucagon (Gcg) at postnatal day (P) 0 in control and Lsd1^Δpan mice (representative images, n = 6 embryos per genotype). Boxed areas are shown in higher magnification. Scale bar, 50 µm. c Quantification of pancreatic epithelial area at e12.5 and e15.5 and Ngn3⁺ cells relative to epithelial area. Data are shown as means ± S.E.M. (n = 3 embryos per genotype; source data are provided as a Source Data file). P = 0.65, 0.02, and 0.39, respectively, Student’s t-test, 2 sided. d Immunofluorescent staining for Ins with somatostatin (Sst), pancreatic polypeptide (Ppy) and ghrelin (Ghrl) at P0 in control and Lsd1^Δpan mice. Scale bar, 25 µm. e Strategy for tamoxifen-inducible Lsd1 deletion in embryonic pancreatic progenitors of mice at e10.5 (Lsd1^Δearly) and e12.5 (Lsd1^Δlate). Yellow boxes: exons; green triangles: loxP sites. f Immunofluorescent staining for Lsd1, Ins and Gcg at e18.5 in control, Lsd1^Δearly and Lsd1^Δlate mice (representative images, n = 3 embryos per genotype). Boxed areas are shown in higher magnification. Scale bar, 50 µm. g Immunofluorescent staining for Lsd1 and EpCAM (left panels) and RNAscope in situ hybridization for HoxA1 (right panels, adjacent sections to left panels) in control and Lsd1^Δearly embryos at e12.5 (representative images, n = 3 Lsd1^Δearly embryos and n = 6 control embryos). Bottom images show higher magnification. Scale bar, 50 µm.