Co-transcriptional genome surveillance by HUSH is coupled to termination machinery

Abstract

The HUSH complex recognizes and silences foreign DNA such as viruses, transposons, and transgenes without prior exposure to its targets. Here, we show that endogenous targets of the HUSH complex fall into two distinct classes based on the presence or absence of H3K9me3. These classes are further distinguished by their transposon content and differential response to the loss of HUSH. A de novo genomic rearrangement at the Sox2 locus induces a switch from H3K9me3-independent to H3K9me3-associated HUSH targeting, resulting in silencing. We further demonstrate that HUSH interacts with the termination factor WDR82 and-via its component MPP8-with nascent RNA. HUSH accumulates at sites of high RNAPII occupancy including long exons and transcription termination sites in a manner dependent on WDR82 and CPSF. Together, our results uncover the functional diversity of HUSH targets and show that this vertebrate-specific complex exploits evolutionarily ancient transcription termination machinery for co-transcriptional chromatin targeting and genome surveillance.

Keywords: CPSF; HUSH complex; MPP8; SETDB1; TASOR; WDR82; gene silencing; heterochromatin; transcription termination; transposable elements.

Figure 1.. Intragenic deletion results in unexpected epigenetic silencing at the Sox2 locus

(A) Schematic diagram of the Sox2 locus and its associated CRISPR-Cas9 screen in mESCs: in WT mESCs, Sox2, and its super-enhancer are separated by a 100 kb gene desert (top). Allele-specific GFP and BFP tags were integrated into the Sox2 gene (middle) and two sgRNA were used to delete the space between Sox2 and its enhancer (bottom). CRISPR screen was conducted by transducing a library of sgRNAs covering 20% of the mouse coding genome and induction of Cas9 expression (bottom). (B) Fluorescence-activated cell sorting (FACS) histogram plots of BFP (left) and GFP (right) in WT (top), Sox2-100kb (center), and Sox2-1kb (bottom) mESCs. (C) FACS 2D plots of BFP (x axis) and GFP (y axis) in Sox2–1kb mESCs following transduction with sgRNA library covering 20% of the mouse coding genes and after induction of Cas9 expression. Marked in red is the gating strategy that was used to sort BFP+/GFP+ and BFP+/GFP− populations, which was followed by deep-sequencing and differential analysis of sgRNAs enrichment. (D) Differential enrichment of sgRNAs between GFP-positive and GFP-negative populations that were sorted in (C) as calculated by DE-Seq. Marked in red aregenes associated with the HUSH complex.

Figure 2.. The HUSH complex is responsible for silencing at the Sox2 short allele

(A) MPP8 western blot in Sox2–1kb reporter (control) and MPP8 knockout (KO) subclones. Membranes were probed with antibodies against MPP8 and β-actin as a loading control. (B) FACS 2D plots of GFP (x axis) and forward scatter (y axis) in WT mESCs, Sox2–1kb reporter (control), and MPP8 KO clones. (C) FACS histogram plots of GFP in Sox2–1kb reporter cells transduced with two non-targeting sgRNAs (control) or TASOR-targeting sgRNAs (TASOR KO), 9 days after induction of Cas9 expression. (D) H3K9me3 ChIP-seq coverage aligned independently to Sox2-GFP (green) or Sox2-BFP (blue) alleles in the Sox2–1kb cell line. (E) Chromatin immunoprecipitation quantitative PCR (ChIP-qPCR) of H3K9me3 in Sox2–1kb reporter. Schematic diagram of the Sox2 locus in the Sox2–1kb reporter and location of primers used for the qPCR (top). Percentage of input recovered for H3K9me3 ChIP in Sox2–1kb reporter control (red) and MPP8 KO clones (blue) with the different primers for qPCR (bottom). (F) Top: schematic diagram of the MPP8 and TASOR endogenous tagging strategy in WT mESCs (termed “AGH” for AID-GFP-HAx3). Bottom: western blots in WT (control) and endogenously tagged mESCs, probing for MPP8 (left), HA (right) and β-actin as a loading control. (G) Browser tracks of TASOR, MPP8 and H3K9me3 ChIP-seq in WT mESCs surrounding the Sox2 locus. Highlighted in yellow are the major ChIP-seq pileups for either MPP8 and TASOR (HUSH peaks) or MPP8-only peaks.

Figure 3.. Two classes of HUSH targets distinguished by H3K9me3, transposon content, and response to HUSH depletion

(A) Heatmaps showing ChIP-seq coverage for MPP8, TASOR, H3K9me3, and SETDB1 at MPP8 peaks. The x axis represents distance from MPP8 ChIP peak in kb. Heatmaps were sorted first by H3K9me3 and then by MPP8 and TASOR signal. Regions with high coverage of both MPP8 and TASOR were classified as HUSH targets (H3K9me3-positive; orange and H3K9me3-negative; light blue), whereas regions with low or no TASOR were delineated as MPP8-only (gray). The number of peaks for each class (n) is indicated. ChIP peak orientation in heatmaps is stranded based on the ratio of positive- and negative-stranded PRO-seq signals. (B) Scatterplots showing all MPP8 and TASOR ChIP-seq peaks as a function of log coverage for MPP8 (x axis) and SETDB1 (y axis) ChIP-seq. Marked in red are peaks overlapping H3K9me3 (left) or TASOR (right). (C) Heatmaps showing the presence and relative position of repetitive elements as annotated by repeat masker, centered and sorted as in (A). The x axis represents distance from MPP8 ChIP peak in kb. (D) Heatmaps showing relative position of transcribed exons as annotated by custom transcriptome models derived from RNA-Seq of WT mESCs, centered and sorted as in (A). The x axis represents distance from MPP8 ChIP peak in kb. (E) Heatmaps showing RNA-seq coverage in WT mESCs, centered and sorted as in (A). The x axis represents distance from MPP8 ChIP peak in kb. (F) Heatmaps showing subtraction of RNA-seq coverage in MPP8-depleted mESCs (treated with auxin for 7 days) from control mESCs, centered and sorted as in (A). The x axis represents distance from MPP8 ChIP peak in kb. (G) Heatmaps showing subtraction of RNA-seq coverage in TASOR-depleted mESCs (treated with auxin for 1 day) from untreated control mESCs, centered and sorted as in (A). The x axis represents distance from MPP8 ChIP peak in kb.

Figure 4.. The HUSH complex interacts with RNA polymerase II termination factors and binds nascent RNA

(A) Log of label-free quantification (LFQ) intensity for HA immunoprecipitation (IP) followed by mass-spectrometry in TASOR-AGH (y axis) and untagged control mESCs (x axis). Colored dots highlight associations of proteins with the HUSH complex (blue) or with RNA polymerase II (red). (B) Same as (A), for MPP8-AGH IP-MS. (C) Western blot showing coIP of MPP8 and WDR82 with TASOR (HA IP) in the TASOR-AGH cell line. Samples were incubated with RNaseA at 2 ng/μL during the IP. Inputs (in), non-bound flowthrough (NB), and HA peptide eluted (E) fractions at shown. (D) Enrichment of MPP8 irCLIP-seq for various genomic features. y axis was calculated by normalizing the number of RT stops to both length and level of transcription, as measured by PRO-seq. Exons were annotated by custom transcriptome models derived from RNA-seq of WT mESCs. (E) Aggregate plot showing MPP8 CLIP-seq normalized read count (y axis) relative to the distance from HUSH or MPP8-only ChIP peaks (x axis). HUSH peaks were divided into H3K9me3-positive (blue), H3K9me3-negative (purple), or MPP8-only regions based on the classification in Figure 3A. (F) Heatmaps showing ChIP-seq coverage for RNA polymerase II using different antibodies that recognize various C-terminal domain (CTD) modifications, centered and sorted as in Figure 3A. The x axis represents distance from MPP8 ChIP peak in kb.

Figure 5.. WDR82 is required for HUSH accumulation on chromatin

(A) Heatmap showing odds ratio values for overlap between H3K4me3, H3K9me3, MPP8, TASOR, and WDR82 ChIP-seq peaks. Peaks were called using MACS2 with broad peak settings. Statistics were calculated by BEDTools fisher. The number of peaks for respective proteins is indicated at the bottom. (B) Heatmaps of WDR82 ChIP-seq coverage in WT mESCs, centered and sorted as in Figure 3A. The x axis represents distance from MPP8 ChIP peak in kb. (C) Western blot of WT and TASOR-AGH mESC total cell lysates. Control and WDR82 KO clones from each cell line are shown. Membranes were probed with antibodies against WDR82, L1-ORF1p, and HSP90 as a loading control. (D) Heatmaps showing TASOR (left) or MPP8 (middle) differential ChIP-seq coverage in WDR82 KO – WT backgrounds compared with reanalyzed ChIP-seq coverage for TRIM28 (right). Signal is centered and sorted as in Figure 3A. The x axis represents distance from MPP8 ChIP peak in kb. (E) Aggregate plot showing TASOR ChIP-seq coverage (y axis) relative to the distance from HUSH ChIP peak (x axis) in WT (red) and WDR82 KO (blue) mESCs. (F) Top: browser tracks of TASOR, MPP8 and WDR82 ChIP-seq in WT and WDR82 KO mESCs. Bottom: browser tracks of PRO-seq in WT (orange) and WDR82 KO (blue) mESCs for the positive (+) and negative (−) DNA strands. The arrow marks the observed shifts in ChIP-seq and PRO-seq signals in WDR82 KO samples. (G) Aggregate plot showing TASOR ChIP-seq coverage individually scaled to span (0,1) interval in each sample (y axis) to visualize changes in plot shape, relative to the distance from HUSH ChIP peak (x axis) in WT (red) and WDR82 KO (blue) mESCs. (H) Heatmaps showing subtraction of PRO-seq coverage (WDR82 KO – WT), for sense (S, left) and antisense (AS, right) strands. Heatmaps were centered and sorted as in Figure 3A. The x axis represents distance from MPP8 ChIP peak in kb.

Figure 6.. HUSH is coupled to transcription termination machinery

(A) Western blot in CPSF3-AGH mESCs, either with or without auxin treatment for 6 or 12 hours. Membranes were probed with antibodies against CPSF3 and β-actin. (B) Average plot of CPSF3 ChIP-seq coverage at H3K9me3-negative HUSH peaks in CPSF3-AGH cells with or without auxin represented by blue or red lines, respectively. The x axis indicates distance from HUSH ChIP peak in kb. (C) Average plot of MPP8 ChIP-seq coverage at H3K9me3-negative HUSH peaks in CPSF3-AGH cells with or without auxin represented by blue or red lines, respectively. The x axis indicates distance from HUSH ChIP peak in kb. (D) Fraction of HUSH peaks that overlap GENCODE-annotated transcription termination sites (TTS) for H3K9me3-positive and H3K9me3-negative regions. (E) Aggregate plots of TASOR, MPP8, WDR82, RNAPII (8WG16), and NELF ChIP-seq coverage centered on unique TSS, midpoints, or TTS (based on polyA) for H3K9me3-negative HUSH sites. Plotted are the median (solid line) and 95% CI of 1,000× bootstrap. The ordinate represents genomic distance relative to the feature in kb and the abscissa represents coverage (reads per billion per base per region).

Figure 7.. Model for co-transcriptional, termination machinery-coupled targeting of HUSH to endogenous genic and transposon targets

(A) Schematic model for the mechanism of HUSH binding to endogenous, non-TE gene targets absent H3K9me3 (top-left), and silencing of young, transcriptionally-active TEs (top-right). HUSH associates with termination factors WDR82 and CPSF and tracks with RNAPII to survey transcribed portion of the genome. Loss of WDR82 results in transcriptional readthrough by RNAPII and prevents stable interaction of HUSH with chromatin at both H3K9me3-positive and -negative HUSH targets. We hypothesize that recognition of a cryptic TTS in an L1 or at the Sox2–1kb allele may provide a trigger H3K9me3. (B) Summary of the different regions and relevant features that are associated with H3K9me3 and/or HUSH ChIP-seq signals.