B cell super-enhancers and regulatory clusters recruit AID tumorigenic activity

Abstract

The antibody gene mutator activation-induced cytidine deaminase (AID) promiscuously damages oncogenes, leading to chromosomal translocations and tumorigenesis. Why nonimmunoglobulin loci are susceptible to AID activity is unknown. Here, we study AID-mediated lesions in the context of nuclear architecture and the B cell regulome. We show that AID targets are not randomly distributed across the genome but are predominantly grouped within super-enhancers and regulatory clusters. Unexpectedly, in these domains, AID deaminates active promoters and eRNA(+) enhancers interconnected in some instances over megabases of linear chromatin. Using genome editing, we demonstrate that 3D-linked targets cooperate to recruit AID-mediated breaks. Furthermore, a comparison of hypermutation in mouse B cells, AID-induced kataegis in human lymphomas, and translocations in MEFs reveals that AID damages different genes in different cell types. Yet, in all cases, the targets are predominantly associated with topological complex, highly transcribed super-enhancers, demonstrating that these compartments are key mediators of AID recruitment.

Figure 1. AID Damages Enhancer DNA

(A) Strategy to reveal AID-mediated breaks. In 53BP1^−/− cells DNA lesions at AID off-targets (e.g., Cd83) in G1 are resected in S and G2M by HR repair nucleases, leading to asymmetric RPA binding that can be detected by ChIP-Seq. (B) The visualization of RPA-Seq was improved by plotting the difference in ChIP signals between + and − strands. An algorithm was developed to efficiently detect asymmetric RPA occupancy. The new approach reveals two additional AID targets at the Bcl11a locus that overlap with enhancer elements (highlighted with red asterisks). The nontargeted enhancer is marked with a blue asterisk. DNaseI, RNA (GRO-seq) (Chiarle et al., 2011), and RPA control (53BP1^−/−AID^−/−) tracks are provided. See also Figure S1 and Table S1A.

Figure 2. Tethering and Compartmentalization of AID Targets in the Mouse Genome

(A) AID targets are largely found within A compartments (black upper track) as defined by Hi-C. Red dots identify the location of damaged loci within the genomic domain. The Hi-C data was obtained from pro-B cells. All other experiments involving mouse B cells in the manuscript were done with activated B cells. (B) Circos plot shows the genome-wide distribution of AID targets that are either tethered within regulatory clusters (red dots) or isolated (black dots). (C) Upper: heat map of cis-interaction frequencies revealing TADs within the domain chr4: 42, 683, 983-48,696,419. Lower: Pax5 gene regulatory cluster, as defined by PolII long-range interactions. The targeted promoter is associated with nondamaged (blue asterisks) and damaged (red asterisk) enhancers. DNaseI hypersensitivity, RNA, hypermutation, and chromosomal translocations (TC-Seq) are also shown. The number of interactions is provided above the ChIA-PET links. See also Figure S1.

Figure 3. AID-Targeted Regulatory Clusters Are Predominantly Associated with B Cell SEs

(A) AID activity at the Igκ locus occurs within a 65 kb SE domain displaying long-range chromatin interactions. PolII interactions, RPA, RNA, and H3K27Ac profiles are provided. (B) Venn diagram showing the fraction of AID targets associated with B cell SEs. (C) Example of AID off-targeted SEs at the Aicda-Apobec1 TAD in chromosome 6. (D) H3K27Ac signal at targeted and nontargeted SEs. Igμ (blue, chr12:114640978-114669901), Igκ (magenta, chr6:70659188-70724456), and Igλ (green, chr.16:19002804-19067747) SEs are highlighted. (E) Size distribution of total constituent enhancers in targeted (red line) or nontargeted (black line) SEs. (F and G) Box plots showing the absolute expression or PolII-mediated connections at targeted (red) and nontargeted (open) SEs. Data are represented as the mean ± SEM. See also Figure S2.

Figure 4. Defining Features of Targeted Enhancers

(A) Myc locus showing the distribution of SEs (H3K27Ac-Seq), enhancers (DNaseI-Seq), PolII long-range interactions (ChIA-PET), AID-mediated damage (RPA-Seq), and RNA synthesis (GRO-Seq). AID-targeted enhancers are denoted with red asterisks. (B–F) Box plots comparing the extent of protein recruitment (B), DNaseI-Seq, eRNA synthesis (C), GRO-Seq, PolII interactions (D), PETs, Spt5 (E), and CTCF occupancy (F) at targeted (red boxes) and nontargeted (open boxes) enhancers. Data are represented as the mean ± SEM. See also Figure S3.

Figure 5. Tethered Regulatory Elements Cooperate to Recruit AID Activity

(A) Il4ra, Il21r, and Nsmce1 form a promoter-gene cluster on mouse chromosome 7. Long-range interactions, DNA damage, and SEs are shown. Il4ra^u/u knockin mice were created by replacing the Il4ra promoter (P, blue arrow) for that of Ubc (red arrow). (B) H3K27Ac in wild-type and knockin mouse B cells. (C) mRNA expression (plotted as rpkm values). (D) Hypermutation frequency at Ubc, Il4ra, Il21r, Nsmce1, Myc, Mir142, and Pim1 genes was measured in Il4ra^+/+ (blue bars) and Il4ra^u/u (red bars) activated B cells. P values shown were calculated with Student’s t test for triplicates experiments (Ubc, Il21r, Nsmce1) and Fisher’s exact test (Myc, Mir142, Pim1) for single experiments. Hypermutation at Ubc was measured on chromosome 5 in Il4ra^+/+ (blue bar) and on chromosome 7 in Il4ra^u/u (red bar). Data are represented as the mean ± SEM. See also Figures S3 and S4 and Table S1B.

Figure 6. AID Targets in Human Lymphomas Are Associated with Long-Range Chromatin Interactions and SEs

(A and B) The SHM-Seq protocol detects AID-mediated hypermutation in Ramos B cells, including at the IGH (A) and MYC (B) loci. (C) Rainfall plot displaying the distance between neighboring mutations across the genome of a DLBCL primary tumor (#129). Kataegic domains of clustered mutations are depicted with red dots. Some of the genes associated with kataegis are highlighted. (D) Representation of sequence context at positions −2, −1, and +1 flanking mutated Cs in DLBCL or breast cancer kataegis. The average context of Cs in the entire human genome is also shown. (E) Percent overlap between hypermutated genes from Ramos Burkitt’s lymphoma (blue bars) or primary DLBCL (red bars) in SEs (left), PolII long-range interactions (middle), or mouse AID targets (right). (F) AID hypermutation of the BL6 regulatory cluster in Ramos cells. SEs, PolII long-range interactions, and hypermutation are provided. See also Figure S5 and Tables S1C and S1D.

Figure 7. AID Damages Different Genes in Different Cell Types

(A) Circos diagram showing hotspots of AID-dependent chromosome translocations to Myc^I-SceI in MEFs and B cells. Hotspots only present in B cells (blue lines), MEFs (red lines), or both cell types (green lines) are provided. (B) Overlap of AID targets in MEFs (red bars) or B cells (blue bars) with SEs. (C) Myc translocations to Flnb are primarily detected in MEFs (red bars), where the gene is associated with a SE domain. Conversely, a single translocation is detected in B cells (black bar). See also Figures S6 and S7 and Tables S1E and S1F.