Bystander activation across a TAD boundary supports a cohesin-dependent transcription cluster model for enhancer function

Abstract

Mammalian enhancers can regulate genes over large genomic distances, often skipping over other genes. Despite this, precise developmental regulation suggests that mechanisms exist to ensure enhancers only activate their correct targets. Sculpting of three-dimensional chromosome organization through cohesin-dependent loop extrusion is thought to be important for facilitating and constraining enhancer action. The boundaries of topologically associating domains (TADs) are thought to prevent enhancers acting on genes in adjacent TADs. However, there are examples where enhancers appear to act across TAD boundaries, but it has remained unclear whether a single enhancer can simultaneously activate genes in different TADs. Here we show that some Shh enhancers can activate transcription concurrently not only at Shh but also at Mnx1 located in an adjacent TAD. This occurs in the context of a chromatin conformation maintaining genes and enhancers in close proximity and is influenced by cohesin. To our knowledge, this is the first report of two endogenous mammalian genes transcribed concurrently under the control of the same enhancer and across a TAD boundary. These findings have implications for understanding the design rules of gene regulatory landscapes and are consistent with a transcription cluster model of enhancer-promoter communication.

Keywords: 3D genome; CTCF; DNA-FISH; RNA-FISH; cohesin; loop extrusion; sonic hedgehog; transcriptional hubs.

Figure 1.

Long-range Shh enhancers can activate transcription at Mnx1 in Shh-expressing tissues. (A) Hi-C heat map, created using HiGlass, of the Shh TAD from wild-type mESCs, at 16 kb resolution. Data are from Boyle et al. (2020). Genes, Shh and Mnx1 enhancers, 35 kb deletion, binding locations of DNA-FISH fosmid probes, position of the TALE target, and the CTCF ChIP-seq track, are shown below the heat map. Genome coordinates (in megabases): mm9 assembly of the mouse genome. (B) Mouse embryo cartoon (left) indicating the position and orientation of the tissue sections analyzed (right). Highlighted are the zone of polarizing activity (ZPA; purple), floorplate (fp; blue), pre-motor neurons (pmn; green), foregut (fg; orange), lung buds (l; light blue), and nonexpressing limb and neural tube tissue (red boxes). (C) Representative images of tissues (top) and nuclei (bottom) showing RNA-FISH signal at Shh (cyan) and Mnx1 (red) in the floorplate (fp) and pre-motor neurons (pmn) of the neural tube. (D) Dorsal, (V) ventral. Scale bars, 5 µm. (D) As in C but in the ZPA of the posterior forelimb bud. (E) The percentage of alleles with Shh and Mnx1 RNA-FISH signal in wild-type (wt) and 35 kb deletion (35 del) E11.5 mouse embryos in expressing tissues of the neural tube (floorplate [fp] for Shh and pre-motor neurons [pmn] for Mnx1) and the posterior limb bud (ZPA for both genes). The data were compared using a two-sided Fisher's exact test. (n.s.) Not significant, (**) P < 0.01. The number of alleles scored (n) is shown below. Data for a biological replicate are shown in Supplemental Figure S1A. Proportions transcribed and statistical data are shown in Supplemental Table S1. (F,G) Representative images of tissues and nuclei showing RNA-FISH signal for Shh (cyan) and Mnx1 (red) in the ventral anterior foregut (F) and lung bud (G). Scale bars, 5 µm. (H) The percentage of Shh-expressing alleles in the floorplate (fp) and Mnx1-expressing alleles in the pmns in comparison with expression of both genes in the ZPA, foregut (fg), and lung buds (l) of an E10.5 embryo assayed by RNA-FISH. Data for a biological replicate are shown in Supplemental Figure S1B. The number of alleles scored (n) is shown below.

Figure 2.

Adjacent RNA-FISH signals signify concurrent transcription at Shh and Mnx1 driven by long-range enhancers. (A) Representative image of foregut nuclei from E10.5 embryos showing RNA-FISH signal at Shh (cyan) and Mnx1 (red). Scale bars, 5 µm. (B) Bar graph showing the percentage of active alleles transcribing at both Shh and Mnx1 (brown) or at Shh (gray) or Mnx1 (red) alone in the ZPA, ventral foregut, and lung bud epithelial cells of a wild-type (wt) E10.5 embryo (left) and in the ZPA, pharynx, palate, telencephalon, and diencephalon of wt and 35 del E11.5 embryos (right). Coactivation at both genes versus activation of a single gene at expressing alleles of the wt and 35 del E11.5 embryos was compared using a two-sided Fisher's exact test. (*) P ≤ 0.05 and P > 0.01, (****) P < 0.0001. Data from a biological replicate are shown in Supplemental Figure S2A. Coactivation proportions in E11.5 wild-type and 35 del cells and statistical data are shown in Supplemental Table S3. The proportion of transcribed alleles where either gene alone or both together were detected and statistical analysis on the significance of coactivation by a single enhancer for E10.5 and E11.5 tissues are shown in Supplemental Table S4. (C, top) Schematic of the TALE-Vp64 construct used to target the ZRS (tZRS-Vp64) enhancer. (NLS) Nuclear localization sequence, (2A) self-cleaving 2A peptide. Repeat variable diresidue (RVD) code is displayed at the right using the one-letter amino acid abbreviations. Equivalent TALE-Δ constructs lack the VP64 module. (Bottom) Representative images of mESC nuclei showing RNA-FISH signal for Shh (cyan) only, Mnx1 (red) only, and both at the same allele. Scale bars, 5 µm. (D) The percentage of Shh transcribing and Mnx1 transcribing alleles in wt (left) and 35 del (right) mESCs activated from ZRS targeted by either tZRS-Vp64 or tZRS-Δ. The data were compared using a two-sided Fisher's exact test. (n.s.) Not significant; (**) P < 0.01, (***) P < 0.001, (****) P < 0.0001. Data from a biological replicate are shown in Supplemental Figure S2B. The number of alleles scored, proportions transcribed, and statistical data are shown in Supplemental Table S5. (E) As in B but for wt and 35 del mESCs transfected with tZRS-Vp64. Data from a biological replicate are shown in Supplemental Figure S2C. Statistical data comparing wt and 35 del are shown in Supplemental Table S3. Statistical analysis of the frequency of concurrent activation by tZRS-Vp64 of Shh and Mnx1 on the same allele versus different alleles in wt and 35 del mESCs is shown in Supplemental Table S4.

Figure 3.

Spatial proximity of Mnx1 to ZRS and Shh is optimal for limb bud ZPA transcription. (A, top) Representative image of ZPA nuclei from E10.5 embryos showing DNA-FISH signal for Shh (green), ZRS (red), and Mnx1 (white). (Bottom) Representative nuclei showing Mnx1 nontranscribing (left) and transcribing (right) alleles. Scale bars, 5 µm. (B) Scatter plots showing interprobe distances between each of the two probe pairs indicated on the X- and Y-axes, with the separation between the third pair indicated by the color (in the color bar) in ZPA cells at nontranscribing, Shh transcribing, Mnx1 transcribing, and Mnx1 and Shh transcribing alleles. Dashed red lines indicate alleles where Shh–Mnx1 or Mnx1–ZRS interprobe distances are <350 nm. (C) Bar plots providing categorical analysis of the spatial relationship of Shh, ZRS, and Mnx1 in ZPA cells at nontranscribing, Shh transcribing, Mnx1 transcribing, and Mnx1 and Shh transcribing alleles. Categories are as follows: <350 nm apart, the upper (75%) quartile distance of all Shh transcribing alleles (Supplemental Figure S3B), and >350 nm. Differences between nonexpressing and expressing alleles were identified using Fisher's exact test. (n.s.) Not significant, (*) P ≤ 0.05 and P > 0.01, (**) P ≤ 0.01, (****) P ≤ 0.0001. Data from two biological replicates are combined in B and C. Median and interquartile data for each of the two biological replicates are shown in Supplemental Figure S3C. Statistical analysis of the significance of spatial proximity at transcribing versus nontranscribing alleles and the proportion of interprobe distances <350 nm are shown in Supplemental Tables S6 and S7. (D, top) Representative image of neural tube pre-motor neuron nuclei from E10.5 embryos showing DNA-FISH signal for Shh (green), ZRS (red), and Mnx1 (white). (Bottom) Images of a ZPA nucleus (left) and a pmn nucleus (right). Scale bars, 5 µm. (E) As in B but for pmns and ZPA cells at non-Mnx1 transcribing and all Mnx1 transcribing alleles. (F) As in C but in pmns and ZPA cells at non-Mnx1 transcribing and Mnx1 transcribing alleles. Differences between nontranscribing and transcribing alleles were identified using Fisher's exact test. (n.s.) Not significant, (*) P ≤ 0.05 and P > 0.01, (**) P ≤ 0.01, (****) P ≤ 0.0001. All data are from both biological replicates combined. Median and interquartile data for each of the two biological replicates are shown in Supplemental Figure S3D. Statistical analysis of the significance of spatial proximity at pmn Mnx1 transcribing alleles versus pmn nontranscribing alleles and Mnx1 transcribing and nontranscribing alleles in ZPA cells and the proportion of interprobe distances <350 nm are shown in Supplemental Tables S6 and S7.

Figure 4.

Cohesin optimizes Mnx1 activation across the Shh TAD boundary. (A, top) Time course of TALE transfection and auxin treatment. (Botom) The percentage of Shh transcribing and Mnx1 transcribing alleles in TALE transfected CTCF-AID cells (left) and SCC1-AID cells (right) either untreated (−auxin) or treated with 24 h (CTCF-AID) or 6 h (SCC1-AID) of auxin (+auxin) mESCs activated from the ZRS enhancer targeted by tZRS-Vp64. Data were compared using a two-sided Fisher's exact test. (n.s.) Not significant, (*) P ≤ 0.05 and P > 0.01, (****) P < 0.0001. Data from a biological replicate are shown in Supplemental Figure S4A. Values for the number of alleles scored and statistical evaluation are summarized in Supplemental Table S8. (B) Representative images of untreated (−auxin; left) and treated (+auxin; right) SCC1-AID nuclei showing DNA-FISH probe signal for Shh (green), ZRS (red), and Mnx1 (white). Scale bars, 5 µm. (C) Scatter plots showing interprobe distances between each of the two probe pairs indicated on the X- and Y-axes, with the separation between the third pair indicated by the color (in the color bar) in untreated (−auxin; left) and treated (+auxin; right) SCC1-AID cells. Dashed red lines indicate alleles where Shh–Mnx1 or Mnx1–ZRS interprobe distances are <350 nm. (D) Bar plots providing categorical analysis of the spatial relationship of Shh, ZRS, and Mnx1 in SCC1-AID mESCs with (+) or without (−) auxin. Categories are as follows: <350 nm and >350 nm. Differences between cells with (+) or without (−) auxin were identified using Fisher's exact test. (n.s.) Not significant, (****) P ≤ 0.0001. Data from two biological replicates are combined in C and D. Median and interquartile data for each of the two biological replicates are shown in Supplemental Figure S4C. Statistical analysis of the significance of spatial proximity at transcribing versus nontranscribing alleles and the proportion of interprobe distances <350 nm are shown in Supplemental Tables S9 and S10.

Figure 5.

Cohesin and the ZRS sphere of influence. (Left) Cohesin-mediated chromatin compaction of the Shh TAD enables regulatory signals from the ZRS enhancer to reach and activate Shh >850 kb away and also enables some activation at Mnx1, which is much closer to ZRS genomically but located at the other side of a TAD boundary. (Right) Acute loss of cohesin results in a decompaction of the chromatin domain, putting Shh beyond the reach of signals from ZRS and considerably reducing the opportunity for Mnx1 to be within the enhancer's “sphere of influence”. (Adapted from Gabriele et al. 2022. Adapted with permission from AAAS.)