Skip to main content
. 2020 Jan 28;9:e50670. doi: 10.7554/eLife.50670

Figure 1. Hedgehog signaling regulates acetylation of H3K27 at a subset of GLI binding regions.

(A) Pipeline for identifying different categories of GLI bound regions (GBRs). (B) Heatmap depicting differential H3K27ac enrichment in WT over Shh-/- limb buds for HH-responsive and Stable GBRs. (C) Classification of GBR categories from E10.5 GBRs with H3K27ac in WT limbs. (D-F). H3K27ac enrichment in WT and Shh -/- is shown across a representative genomic region near a Stable GBR (D), and biologically validated HH-responsive GBRs: a HH-dependent GBRs, GRE1, at the HH target gene Gremlin 1 (Grem1) (Li et al., 2014) (E) and HH-sensitive GBRs shown to regulate limb-specific expression of the HH target Ptch1 (Lopez-Rios et al., 2014) (F). (G) HH-dependent GBRs, HH-responsive GBRs and Stable GBRs are significantly enriched (2 kb upstream- 1 kb downstream of TSS) near HH target genes compared to randomly chosen genes (p=0, p=0 and p=0, respectively, permutation test based on 1000 permutations). (H) Proportional distribution of Stable and HH-responsive GBRs arounds transcription start sites (TSS), indicating significant enrichment of Stable GBRs at TSS compared to HH-responsive GBRs (p=2.55e-40, Fisher's exact test, two sided). (I) Both HH-dependent and HH-sensitive GBRs have significantly more GLI motifs than Stable GBRs (top)(p=2.2e-16 and p=8.00e-06; one-sided proportional test). HH-dependent and HH-sensitive GBRs containing GLI motifs have significantly higher quality of GLI motifs than Stable GBRs (Quality score; p=5.03e-13 and p=5.98e-08; one-sided Wilcoxon test) and significantly more motifs per GBR within HH-dependent GBRs than Stable GBRs (Quantity score; p=5.92e-06; one-sided Wilcoxon test). See Figure 1—figure supplement 1, Figure 1—source data 1, Figure 1—source data 2, Figure 1—source data 3, Figure 1—source data 4.

Figure 1—source data 1. Endogenous GLI3-Flag ChIP-seq analyzed data and called peaks.
GLI3 binding regions with called peaks with a false discovery rate (FDR) < 0.05 from two biological replicates of E10.5 (32–35S) forelimbs. Rank ordered coordinates, peak length, log2 fold change (log2FC) and FDR are listed for each peak.
elife-50670-fig1-data1.xlsx (346.8KB, xlsx)
Figure 1—source data 2. WT vs Shh-/- H3K27ac ChIP-seq analyzed data and called peaks.
H3K27ac called peaks with a FDR < 0.05 from two biological replicates from WT and Shh-/- E10.5 forelimbs. For each peak, the assigned Peak ID, coordinates, peak type, fold change normalized to input for WT and Shh-/- samples and fold change of WT over Shh-/- are listed. Additional tabs include sorted datasets for sub-classifications. Tabs containing GBRs indicate intersections with GLI binding regions.
Figure 1—source data 3. H3K4me1 ChIP-seq analyzed data and called peaks from GSE86690.
H3K4me1 called peaks with a false discovery rate (FDR) < 0.05 from two biological replicates of E10.5 WT forelimbs. Note that this is a reanalysis of a publicly available ENCODE dataset (see methods).
Figure 1—source data 4. Motifs uncovered from HH-responsive enhancers.
Table showing the top 20 motifs uncovered from de novo motif analysis on HH-responsive GBRs. The enrichment is relative to matched genomic controls. Note that ‘HH_resp_2’ is the only motif with an enrichment value of greater than two and corresponds with a known GLI binding motif.

Figure 1.

Figure 1—figure supplement 1. Nuclear localization of GLI3 and properties of GLI binding regions.

Figure 1—figure supplement 1.

(A) Intersection of endogenous GLI3 binding and H3K27ac in E10.5 WT limb buds. (B) Western blots from anterior and posterior E10.5 limb buds indicating the distribution of endogenous GLI3-FLAG in cytoplasmic and nuclear fractions (C = cytoplasmic fraction, N = nuclear fraction; Ant = Anterior forelimb, Post = Posterior forelimb) (n = 3). (C) Hedgehog-responsive enhancers that are not bound by GLI are clustered near GLI binding regions. Box plot indicates the proximity of HH-responsive H3K27ac peaks that are not bound by GLI to either HH-Responsive GBRs or Stable GBRs compared to random peaks. For both HH-responsive and stable GBRs, the number of HH-Responsive non-GBR H3K27ac peaks is significantly larger than the number of random regions (Wilcoxon-test p-value=0). (D) HH-responsive peaks not bound by GLI3 are clustered together. The genome was split into 100,000 base-pair non-overlapping windows and the number of HH-responsive H3K27ac peaks that are not bound by GLI3 were counted as well as the number of random peaks. Only windows that overlapped with at least one HH-responsive H3K27ac peak or random peak were considered. The two counts are significantly different (Wilcoxon-test p-value=0). The dark black line indicates the median. The lower boundary of the box indicates the first quantile, while the upper boundary of the third box is the third quantile. The circles indicate outliers. (E) Box plot showing the conservation scores for different classes of GBRs. The conservation scores correspond to phastCons values linearly scaled from 0 to 255. HH-responsive GBRs have significantly lower conservation scores than stable GBRs (p-value=0.0001134492, one sided Wilcoxon test). None of the other pairs of GBRs are significantly different from each other. ‘Coding regions’ represent conservation scores for all protein coding genes in the mouse mm10 genome while ‘Random regions’ represent conservation scores for a set of 1000 random genomic loci that do not overlap with any gene. The dark black line indicates the median. The lower boundary of the box indicates the first quantile, while the upper boundary of the third box is the third quantile. The circles indicate outliers.