LDB1 establishes multi-enhancer networks to regulate gene expression

Abstract

How specific enhancer-promoter pairing is established remains mostly unclear. Besides the CTCF/cohesin machinery, few nuclear factors have been studied for a direct role in physically connecting regulatory elements. Using a murine erythroid cell model, we show via acute degradation experiments that LDB1 directly and broadly promotes connectivity among regulatory elements. Most LDB1-mediated contacts, even those spanning hundreds of kb, can form in the absence of CTCF, cohesin, or YY1 as determined using multiple degron systems. Moreover, an engineered LDB1-driven chromatin loop is cohesin independent. Cohesin-driven loop extrusion does not stall at LDB1-occupied sites but aids the formation of a subset of LDB1-anchored loops. Leveraging the dynamic reorganization of nuclear architecture during the transition from mitosis to G1 phase, we observe that loop formation and de novo LDB1 occupancy correlate and can occur independently of structural loops. Tri-C and Region Capture Micro-C reveal that LDB1 organizes multi-enhancer networks to activate transcription. These findings establish LDB1 as a driver of spatial connectivity.

Keywords: chromatin architecture, enhancer, LDB1, looping, CTCF, cohesin, YY1, cell cycle, hub, LMO2.

Figure 1.. LDB1 mediates chromatin contacts between cis-regulatory elements.

(A) Numbers of structural loops and CRE loops that are weakened, unchanged or strengthened upon LDB1 depletion. Data shown for structural loops with LDB1 at 0 or 1 anchors. (B) Distribution of weakened CRE loop types. Fraction of loops with RAD21/CTCF co-occupied peaks in both anchors (below). (C) Fraction of enhancers and promoters in G1E-ER4 cells occupied by LDB1, YY1 and CTCF. (D) Schematic representing the motif analysis strategy for heterotypic loops and the top 10 most enriched motifs. (E) Change in loop strength upon LDB1 depletion for loops categorized based on LDB1 and CTCF occupancy. P-values calculated using a two-sided Mann-Whitney U test. (F) LDB1-dependent homotypic loop (red arrow) and LDB1-dependent heterotypic loop (green arrow).

Figure 2.. LDB1-dependent CRE loops are associated with transcription activation.

(A) Gene expression changes measured by TT-seq upon LDB1 depletion (n=3). (B) Gene expression changes (TT-seq) for genes categorized by the number of loop anchors overlapping their TSS. P-values calculated using a two-sided Mann-Whitney U test. (C) Baseline gene expression measured by TT-seq. Genes categorized by the number of LDB1 dependent or independent CRE loops they interact with. (D) Cumulative frequency distributions for gene distance to nearest LDB1 ChIP-seq peak. (E) Numbers of inter-TAD vs intra-TAD LDB1-dependent CRE loops. (F) Loop lengths for LDB1-dependent inter-TAD and intra-TAD CRE loops. P-values calculated using a two-sided Mann-Whitney U test. (G) Loop strengths for LDB1-dependent inter-TAD and intra-TAD CRE loops. Loop strength calculated using 5k resolution; P-values calculated using a two-sided Mann-Whitney U test.

Figure 3.. LDB1 forms fine-scale looped networks at LDB1-dependent genes

(A) Numbers of LDB1-dependent loops detected by Micro-C or RCMC. (B) Proportions of LDB1 or CTCF ChIP-seq peaks overlapping weakened loop anchors. (C) Examples of LDB1-dependent looped networks. (D) 10k resolution TRI-C contact maps for MYC proximal and distal regions. Contacts represent multi-way interactions involving the MYC promoter.

Figure 4.. LDB1 occupancy is mutually independent of YY1, CTCF and cohesin at most locations.

(A) LDB1 ChIP-seq peaks that are occupied by cohesin (RAD21), YY1, or CTCF. (B) LDB1-occupied enhancer elements that are occupied by cohesin (RAD21), YY1, or CTCF. (C) ChIP-seq profiles in LDB1-AID cells for RAD21, CTCF and YY1 before/after LDB1 depletion. (D) RAD21 ChIP-seq signal at RAD21 ChIP-seq peaks overlapping CTCF peaks or LDB1 peaks. P-values calculated using a two-sided Mann-Whitney U test. (E) ChIP-seq profiles in SMC3-AID, CTCF-AID, and YY1-AID cells before/after 4hr auxin treatment.

Figure 5.. LDB1 Can Function in the absence of cohesin

A. Change in loop strength for LDB1-dependent CRE loops in response to LDB1 depletion (darker colors) or SMC3 depletion (lighter colors). P-values calculated using a Wilcoxon signed-rank test. 3 LDB1 dependent loops were in sparse regions in SMC3-AID datasets and removed from analysis. B. H3K27ac ChIP-seq signal at enhancers within LDB1-only loop anchors or dual-sensitive loop anchors. P-values calculated using a two-sided Mann-Whitney U test. C. Relative RNA levels for β-globin measured by RT-qPCR in SMC3-AID cells −/+ ZF-SA and −/+ auxin (4hr). P-values calculated using One-way ANOVA. Data are represented as mean ± Std Dev. D. Lengths of LDB1 only and LDB1/cohesin dual sensitive loops. E. Distance to encompassing structural loop anchors for LDB1-only loops and dually sensitive loops. Only loops with an encompassing structural loop are shown. P-values calculated using a two-sided Mann-Whitney U test. F. APA plots for LDB1-dependent CRE loops stratified by their response to SMC3 depletion and whether they are encompassed by a structural loop. Numbers represent raw center pixel values.

Figure 6.. LDB1 chromatin occupancy correlates with loop establishment during G1-phase entry

(A) ChIP-seq profiles for LDB1 at each cell cycle stage at all LDB1 peaks identified in asynchronous cells. (B) APA plots from 10k resolution Hi-C data at each cell cycle stage for each category of LDB1-dependent CRE loops. Average ChIP-seq profiles are shown for each loop type for LDB1 peaks within loop anchors. (C) Loop strength (top) and observed contacts between loop anchors (bottom) for each category of LDB1-dependent CRE loops and for structural loops at each cell cycle stage. Median loop strength and observed contacts normalized to prometaphase are shown for each loop category. (D) Examples of an LDB1/cohesin dually sensitive loop and LDB1-only loop. Green arrow indicates the LDB1-dependent loop, blue arrow indicates an encompassing structural loop.