An RB-EZH2 Complex Mediates Silencing of Repetitive DNA Sequences

Abstract

Repetitive genomic regions include tandem sequence repeats and interspersed repeats, such as endogenous retroviruses and LINE-1 elements. Repressive heterochromatin domains silence expression of these sequences through mechanisms that remain poorly understood. Here, we present evidence that the retinoblastoma protein (pRB) utilizes a cell-cycle-independent interaction with E2F1 to recruit enhancer of zeste homolog 2 (EZH2) to diverse repeat sequences. These include simple repeats, satellites, LINEs, and endogenous retroviruses as well as transposon fragments. We generated a mutant mouse strain carrying an F832A mutation in Rb1 that is defective for recruitment to repetitive sequences. Loss of pRB-EZH2 complexes from repeats disperses H3K27me3 from these genomic locations and permits repeat expression. Consistent with maintenance of H3K27me3 at the Hox clusters, these mice are developmentally normal. However, susceptibility to lymphoma suggests that pRB-EZH2 recruitment to repetitive elements may be cancer relevant.

Keywords: H3K27me3; Polycomb; cancer; epigenetics; heterochromatin; repetitive DNA; retinoblastoma protein.

Figure 1. pRB Associates with Genomic Repeats in Murine and Human Fibroblasts

(A) pRB western blots of MEF chromatin fractions. Coomassie-stained histones indicate relative chromatin quantities. (B) Overall genomic distribution of pRB ChIP-seq peaks per growth condition; n = 424,588 peaks and n = 77,809 peaks for mouse pRB in arrested and proliferating MEFs respectively; n = 71,511 peaks for human pRB in arrested IMR90 cells. (C) Heatmaps display scaled pRB ChIP-seq read buildups at proximal promoter regions occupied by wild-type pRB peaks. Each row contains ± 1 kb of flanking sequence. The intensity scale indicates the read enrichment level. (D) Percent distribution of pRB ChIP-seq peaks per repeat class; n = 321,892 peaks for mouse pRB arrested, n = 49,210 peaks for mouse pRB proliferating, and n = 99,186 peaks for human pRB arrested. (E) Genome viewer tracks display mouse pRB ChIP-seq reads at Ccne2, with genomic coordinates above. Repeat Masker and RefSeq tracks are shown below. Red bars denote peaks. (F) The analogous region of human CCNE2 is shown and labeled akin to (D). (G) Example of mouse pRB occupancy at a LINE-1 element 3′ of the Ccne2 gene. (H) Example of a human pRB peak 5′ of CCNE2 that simultaneously overlaps multiple repeat elements.

Figure 2. Cell-Cycle-Independent pRB-Repeat Association

(A) Western blots of whole-cell extracts display expression levels of wild-type and the F832A mutant (Rb1^S) pRB as well as p107 and p130. pRB western blot of chromatin fractions from arrested and proliferating cells. Coomassie-stained histones indicate loading. (B) pRB ChIP-qPCR at the Mcm3 TSS and 2 kb 5′ of Mcm3. (C) Heatmaps of pRB ChIP-seq read enrichment per repClass. Each row represents one scaled wild-type peak location at an element within the repClass and includes ± 1 kb of flanking sequence. Intensity scale indicates magnitude of read enrichment. (D) pRB ChIP-qPCR at the indicated repetitive elements. (E) Representative genomic regions with pRB repeat association at LINE-1 fragments. Red bars denote peaks. For all graphs, error bars indicate one SD from the mean, and an asterisk represents a significant difference from wild type (p % 0.05 by t test). See also Figures S1–S3.

Figure 3. pRB Silences Repetitive Element Expression

(A) Heatmap of repeat expression from three Rb1^S/S MEF preparations relative to the average of three wild-type replicates. RNA-seq reads were aligned to repeat indices, binned according to repClass and repFamily, and normalized to the total number of aligned reads in the sample. Expression was quantified as a log₂ ratio relative to the average of three wild-type replicates. (B) RNA-seq reads aligned to instances of LINE, IAP endogenous retrovirus, and major satellite repeats. (C) qRT-PCR of the indicated repetitive elements in proliferating MEFs plotted as log2 of the ratio, with wild type using actin as an internal control. Each MEF pair was cultured independently three times and expression levels for each replicate are shown to illustrate variability in expression between culture and genotypes. (D) Expression microarrays performed with RNA from arrested MEFs. Log2 values of mutant/wild type shown as a heatmap depict expression levels of endogenous retrovirus-detecting probe sets. See also Figure S4.

Figure 4. H3K27me3 Enrichment at Repetitive Elements Is pRB Dependent

(A) ChIP-qPCR for H3K27me3 in arrested MEFs. Error bars indicate one SD from the mean. An asterisk represents a significant difference from wild type (p ≤ 0.05 using a t test). (B) Overall genomic distribution of H3K27me3 ChIP-seq peaks per growth condition; n = 656,342 peaks in arrested cells; n = 143,252 peaks in proliferating cells. (C) Genome viewer tracks depict H3K27me3 read buildup at L1 elements (highlighted by red boxes), with genomic coordinates and scale above tracks. (D) H3K27me3 distribution at a Hox gene cluster. (E) Heatmaps of H3K27me3 read enrichment per repClass. Each row represents one scaled wild-type peak location at an element within the repClass and includes ± 1 kb of flanking region. Intensity scale indicates magnitude of read enrichment. See also Figures S5 and S6.

Figure 5. pRB-Chromatin Association Mediates EZH2 Recruitment to Repetitive DNA

(A) Overall genomic distribution of EZH2 ChIP-seq peaks in arrested wild-type MEFs; n = 840,543 peaks. (B) Heatmaps of EZH2 read enrichment per repClass. Each row represents one scaled wild-type peak location at an element within the repClass that includes ± 1 kb of flanking region. Intensity scale indicates level of read enrichment. (C) EZH2 ChIP-qPCR at repeats. (D) EZH2-pRB ChIP-reChIP at repeats, with genetic knockouts as controls. (E) Two genomic regions depict ChIP-seq tracks for EZH2, pRB, and H3K27me3 in arrested MEFs, with genomic coordinates and scale above tracks. Red boxes highlight peak overlaps across datasets. For all graphs, error bars indicate one SD from the mean. An asterisk represents a significant difference from wild type (p ≤ 0.05 using a t test). See also Figure S7.

Figure 6. Altered Chromatin and Repeat Expression in Rb1^S/S Splenocytes

(A) Heatmap of repeat element expression per tissue of 6- to 8-week-old mice quantified by qRT-PCR. Log2 ratio of expression is displayed relative to the average of all wild-type measurements for a given element in each tissue. (B) Heatmap of interferon gene expression in splenocytes from the same mice. (C) H3K27me3 and H3 ChIP-qPCR from freshly harvested splenocytes of 6-week-old mice. Error bars indicate one SD from the mean (n = 3; an asterisk indicates p < 0.05, t test).

Figure 7. Aged Rb1^S/S Mice Are Tumor Prone

(A) Kaplan-Meier plot of tumor-free survival for cohorts of mice aged until animal protocol endpoint. Tick marks indicate mice necropsied at intermediate ages. Mutant mice are significantly more cancer prone than wild-type mice (log rank test, p < 0.05). (B) Anatomical location and cancer type listed by frequency of occurrence. (C) Peritoneal cavity images upon necropsy. A white arrow indicates an abnormal mass in the spleen of mouse C6217. A dashed line highlights the mesenteric lymph node of mouse C6209. A white arrow indicates an abnormal liver, and a black arrow indicates a normal lobe in mouse C5360. (D) H&E staining of tissue sections from the abnormality indicated in the mutant animals above. A black arrow indicates the lymph node in mouse C6209. The scale bars represent 300 μm.