GLI transcriptional repression is inert prior to Hedgehog pathway activation

Abstract

The Hedgehog (HH) pathway regulates a spectrum of developmental processes through the transcriptional mediation of GLI proteins. GLI repressors control tissue patterning by preventing sub-threshold activation of HH target genes, presumably even before HH induction, while lack of GLI repression activates most targets. Despite GLI repression being central to HH regulation, it is unknown when it first becomes established in HH-responsive tissues. Here, we investigate whether GLI3 prevents precocious gene expression during limb development. Contrary to current dogma, we find that GLI3 is inert prior to HH signaling. While GLI3 binds to most targets, loss of Gli3 does not increase target gene expression, enhancer acetylation or accessibility, as it does post-HH signaling. Furthermore, GLI repression is established independently of HH signaling, but after its onset. Collectively, these surprising results challenge current GLI pre-patterning models and demonstrate that GLI repression is not a default state for the HH pathway.

Fig. 1. GLl3 binds to poised, accessible chromatin prior to HH signaling.

a Schematic for HH induction timeline during limb development. b Quantification of 21–25 S embryos with expression of Gli1 and Shh assayed by in situ hybridization. c Representative western blot (n = 3) of endogenous GLl3^FLAG protein in the limb bud pre-(21–23 S) and post-HH signaling (32–35 S). d Venn diagram of all pre- and post-HH identified GLl3 CUT&RUN called peaks. e Percentage of E10.5 GLl3-bound regions also bound at E9.25, before HH signaling, for all E10.5 identified GBRs and E10.5 HH-responsive GBRs. f Venn diagram of HH-responsive GBRs enriched for poised enhancer marks and bound by GLl3 at E9.25. Poised enhancer modifications identified by H3K4me1 CUT&Tag (n = 3), H3K4me2 ChlP-seq (n = 2), ATAC-seq (n = 2). g Heatmap of GLl3 enrichment pre­ and post-HH signaling and enrichment of poised enhancer marks pre-HH signaling. Many regions lacking GLl3 also lack enrichment of poised marks. h Example of a HH-responsive GBR (orange shading) that is poised, accessible, and bound by GLl3 prior to HH signaling at a limb-specific Ptch1 enhancer. i Example of a validated HH-responsive GBR GRE1, that regulates the GLl3 target Gremlin, that is inaccessible, lacks H3K4me1 and GLl3 binding at E9.25. Scale bars = 1 kb. Source data are provided as a Source Data file.

Fig. 2. GLl3 binding regions have reduced enrichment of H3K27ac and are enriched with HDACs in pre-HH limb buds.

a H3K27ac ChlP-seq tracks showing a E9.25 (21-23 S) GLl3-bound HH-responsive GBR with reduced acetylation in E9.25 WT limb buds compared to E10.5 WT limbs (note that the levels of enrichment are comparable to E10.5 Shh^−/− limbs). b Scatter plot of H3K27ac enrichment at HH-responsive GBRs in pre- and post-HH WT limbs. HH-responsive GBRs with significant reductions (FDR < 0.05) in H3K27ac at E9.25 compared to E10.5, are denoted in dark blue, while the 3 GBRs with significantly higher H3K27ac in E9.25 limbs are in orange (FDR < 0.05 n = 2). c Heatmap of H3K27ac ChlP-seq indicating relative H3K27ac enrichment at HH responsive GBRs in pre- and post-HH limbs (n = 2). d GLl3, HDAC1 and HDAC2 CUT&RUN tracks at E9.25 GLI3-bound regions, pre- and post-HH. e HDAC localization at E10.5 bound GBRs. Prior to HH signaling, most regions are enriched for both HDAC1 and HDAC2, while after, most GBRs are enriched for only HDAC1 (E9.25 HDAC1/2 n = 2; E10.5 HDAC1/2 n = 3). f Initially most HDAC1-bound regions are also enriched for HDAC2, while after HH-signaling, fewer HDAC1-bound regions are also enriched for HDAC2. g Prior to HH signaling, GLl3-bound regions are enriched for poised enhancer marks with lower levels of H3K27ac and HDAC enrichment. These are comparable to post-HH GLl3 enhancers where GLl3 actively represses targets. Orange shading in tracks indicates HH-responsive GBRs defined in Supplementary Fig. 1e. Scale bars = 1 kb.

Fig. 3. Loss of Gli3 does not result in premature de-repression of enhancers.

a Schematic for testing whether loss of Gli3 prematurely increases H3K27ac levels through GLI3 de-repression. b Heatmap of H3K27ac enrichment at HH-responsive GBRs in pre-HH (E9.25, 21–23 S) and post-HH (E10.5, 33-34 S) limb buds, with loss of Gli3. At E10.5, Shh^−/− limbs have reduced H3K27ac due to GLI3 repression. Loss of Gli3 in Shh^−/−;Gli3^−/− results in de-repression of target enhancers and increased acetylation. This change is not observed prior to HH signaling with loss of Gli3 compared to WT limb buds. c Scatterplot of H3K27ac enrichment in pre-HH WT vs. Gli3^−/− limbs shows no increase in acetylation with loss of Gli3. d H3K27ac ChIP-seq tracks at a pre-HH GLI3-bound HH-responsive GBR with low H3K27ac in E9.25 WT limb buds that does not increase acetylation levels in E9.25 Gli3^−/−. e Examples of regions, bound by GLI3 at E9.25, that do not increase H3K27ac levels with loss of Gli3 in pre-HH limb buds as they do in Shh^−/−;Gli3^−/− post-HH limbs. f Schematic depicting loss of Gli3 in pre-HH limb buds does not lead to de-repression of target enhancers as it does after HH induction at E10.5. Orange shading on tracks indicates HH-responsive GBRs defined in Supplementary Fig. 1e. Scale bars = 1 kb.

Fig. 4. Most genes are not de-repressed in Gli3 mutants prior to HH induction.

a Schematic for testing whether loss of Gli3 can prematurely activate target genes through GLI3 de-repression. b, c Volcano plot of differentially expressed genes (DEGs) detected in RNA-seq between WT and Gli3^−/− limbs E9.25 (21-23 S) (b) and E10.5 (32–35 S) anterior limb buds (c). d Fluorescent in situ hybridization showing maximum intensity projections for Hand2, Foxf1 and Ptch1 in E9.25 WT and Gli3^−/− limb buds indicating the absence of detectable Foxf1 and Ptch1 in pre-HH limb buds. Dashed white lines outline the limb bud region. e In situ hybridization for the GLI3 target Hand2, in WT and Gli3^−/− limbs pre- (21, 23 S) and immediately post-HH induction (24–27 S; n = 3). Note that Hand2 is expressed almost uniformly across the limb bud in 21 S embryos in both WT and Gli3^−/− limbs. At 23 S, slight reduction in anterior expression of Hand2 is observed in both WT and Gli3^−/− limbs (white arrowhead). The red brackets denote anterior limb. Images acquired on a Zeiss LSM 710 Confocal visualized using maximal intensity projections with pseudo-coloring for merged images. f Schematic identifying H3K27me3 enriched regions in E9.25, E10.5 anterior (GLI3-repressed), and E10.5 posterior (HH signaling, loss of GLI-R) limb buds using CUT&Tag. g–j Orange shading in tracks indicates HH-responsive GBRs defined in Supplementary Fig. 1e. g Example of a HH-responsive GBR at a HH target gene promoter with H3K27me3 enrichment specifically at E9.25 but not E10.5. h Representative distal limb-specific HH-responsive GBR downstream of Ptch1 with no H3K27me3 enrichment in E9.25 or E10.5 limb buds. i Examples of HH target genes associated with polycomb repression that could potentially be incompetent for GLl3 repression. j Examples of genes that are bound by GLl3 at E9.25, contain H3K27ac enrichment and lack H3K27me3 enrichment, suggesting they are competent for GLl3 repression but are not de-repressed in the absence of Gli3. For f–j see Supplementary Data 1. Scale bars for tracks indicate 1 kb.

Fig. 5. GLl3-dependent chromatin compaction occurs after HH induction.

a ATAC-seq tracks showing examples of HH-responsive GBRs that are compacted in posterior E10.5 (35 S) Shh^−/− limbs but not in E9.25 (21–23 S) WT or posterior E10 (28–30 S) Shh^−/− limbs. Note posterior E10.5 Shh^−/−;Gli3^−/− limbs maintain accessible chromatin at these regions. b Violin plots of chromatin accessibility in WT and Shh^−/− limbs. Globally the chromatin at HH-responsive GBRs is more accessible in posterior E10 Shh^−/− limb buds compared to posterior E10 WT limbs and posterior E10.5 Shh^−/− limbs. For WT E9.25 (21–23 S) and Shh^−/− E10.5 (35 S), n = 2; for WT and Shh^−/− E10 (28–30 S) and WT E10.5 (35 S), n = 3; for E10.5 (33-35 S) Shh^−/−;Gli3^−/−, n = 5 biologically independent replicates. Two-sided Wilcoxon signed rank tests were performed to compare each pair of groups, multiple hypothesis testing adjusted using BH method. E9.25 < E10 WT, FDR < 2.69E-15; E9.25 < E10 Shh^−/− FDR < 3.44E-23; E9.25 < E10.5 FDR < 5.82E–15 E10 WT < E10 Shh^−/−, FDR < 2.55E-4; E10 Shh^−/− >E10.5 Shh^−/−, FDR < 1.21-18; E10.5 WT > E10.5 Shh^−/−, FDR < 9.81E-12). Red dashed lines indicates E10.5 Shh^−/− mean. c Average ATAC profiles of HH-responsive GBRs in posterior E10 and E10.5 Shh^−/− limb buds. d Violin plots depicting log2 fold changes in chromatin accessibility (WT versus Shh^−/−) at E10 and E10.5 (p = 0.000082 two-sided Wilcoxon signed rank test). e, f Scatter plots of ATAC-seq signal in WT vs. Shh^−/− limbs at E10 e and E10.5 f. GBRs and their associated genes annotated in red signify a significant reduction of chromatin accessibility in Shh^−/− limbs compared to WT (FDR < 0.05). GBRs in blue indicate visibly reduced accessibility in E10 Shh^−/− limb buds that are not significantly changed until E10.5 g ATAC-seq tracks of GBRs near Ptch1 that are reduced, but not significant, in E10 Shh^−/− limbs compared to E10 WT limb buds. h, i GBRs that are already have significantly reduced accessibility in E10 Shh^−/− limbs compared to E10 WT. j Schematic showing the temporal onset of chromatin compaction at GBRs in relation to the initiation of HH signaling. k Quantification of GLl3 present along ciliary axonemes out of total number of cilia (marked by ARL13b) that colocalize with GLl3 in E9.25 (21–23 S) (n = 3), E9.75 (26-28 S) center limb buds (n = 6) and E10.5 (35 S) anterior and posterior limbs (n = 3). Error bars indicate SEM. Unpaired, two-sided t-tests were performed. E9.25 (21–23 S) vs E9.75 (26–28 S), p = 0.0043; E9.75 (26–28 S) vs E10.5 (35 S) anterior, p = 0.00025; E9.25 (21–23 S) vs E10.5 (35 S) anterior, p = 0.048; E9.25 (21–23 S) vs E10.5 (35 S) posterior, p = 0.134; E9.75 (26–28 S) vs E10.5 (35 S) posterior, p = 0.00061. l, m Representative images of GLI3^FLAG ciliary distribution in E9.25 l and E9.75 m limb buds acquired on a Zeiss LSM 710 Confocal visualized using maximal intensity projections with pseudo-coloring for merged images. For E9.25, n = 3 and for E9.75 n = 6 biological replicates on individual embryos. For each replicate, ~75 cilia were analyzed and quantified. n Schematic depicting the temporal shift in GLl3 ciliary localization. Orange shading in all tracks indicates HH-responsive GBRs defined in Supplementary Fig. 1e. Scale bars for ATAC-seq tracks =1 kb. Scale bars for panel l = 0.5 µm. Source data are provided as a Source Data file.

Fig. 6. Model for establishing GLl3 repression.

a Model for lack of GLl3 repression prior to HH signaling. In pre-HH limb buds, many target genes are already activated by HH-independent factors. While GLl3 binds to most targets, it is inert as it is unable to regulate deposition of H3K27ac at enhancers or chromatin accessibility. After HH induction when ‘active’ GLl3 repression has been established, GLl3 prevents addition of H3K27ac at enhancers to spatially restrict expression of its target genes and pattern the limb. b Schematic of GLl3 spatial regulation of target genes. In pre-HH limb buds, GLl3 does not repress targets such as Hand2, which is expressed in an expanded anterior domain. In contrast, GLl3 restricts expression of targets to the posterior limb bud where HH signaling is active in post-HH limbs. c Schematic for GLI repression being established in a context-dependent manner. d Possible models for initiating GLl3 repressor activity. Left: HH-independent co-repressor(s), which may not be abundant in early limb development collaborate with GLl3 to assemble a complete repression complex. Middle: Addition of post-translational modifications (PTMs), potentially added via ciliary trafficking and processing, promote assembly of, or stabilize, a GLI repression complex. Right: A spike in ciliary GLl3 processing, as reflected by increased axonemal colocalization, increases the amount of viable GLl3 repressor, potentially through addition of PTMs as suggested above.