DNA methylation governs the sensitivity of repeats to restriction by the HUSH-MORC2 corepressor

Abstract

The human silencing hub (HUSH) complex binds to transcripts of LINE-1 retrotransposons (L1s) and other genomic repeats, recruiting MORC2 and other effectors to remodel chromatin. How HUSH and MORC2 operate alongside DNA methylation, a central epigenetic regulator of repeat transcription, remains largely unknown. Here we interrogate this relationship in human neural progenitor cells (hNPCs), a somatic model of brain development that tolerates removal of DNA methyltransferase DNMT1. Upon loss of MORC2 or HUSH subunit TASOR in hNPCs, L1s remain silenced by robust promoter methylation. However, genome demethylation and activation of evolutionarily-young L1s attracts MORC2 binding, and simultaneous depletion of DNMT1 and MORC2 causes massive accumulation of L1 transcripts. We identify the same mechanistic hierarchy at pericentromeric α-satellites and clustered protocadherin genes, repetitive elements important for chromosome structure and neurodevelopment respectively. Our data delineate the epigenetic control of repeats in somatic cells, with implications for understanding the vital functions of HUSH-MORC2 in hypomethylated contexts throughout human development.

Fig. 1. Studying the epigenetic control of repeats in a somatic cell model of brain development.

A Simplified model of epigenetic transcriptional regulation of genomic repeats by DNA methylation and the HUSH-MORC2 corepressor. Factors shaded in purple are subject of particular focus in this study. B Brightfield image and immunostaining of the hNPC line used in this study illustrating Nestin expression. Imaging was repeated frequently during the study to confirm cell morphology and marker expression. C Assessment of genome-wide CpG methylation according to whole-genome Oxford Nanopore sequencing of the hNPC line (n = 1) at approximately ×40 coverage. The violin plots represent the median CpG methylation of the given set of genomic intervals and span the interquartile range. D Schematic of CUT&RUN epigenome profiling experiments in human fetal forebrain tissue (n = 2) and hNPCs (n = 3). E, F Heatmaps illustrating CUT&RUN signal enrichment of H3K4me3, H3K9me3 and a non-targeting IgG control in fetal forebrain samples and hNPCs plotted over consensus peaks called in fetal samples. Displayed are the genomic regions spanning ±10-kb from the peak center. G Schematic of CRISPRi experiments to deplete MORC2 or HUSH subunit TASOR. H Genome browser snapshots of MORC2 (left) and TASOR (right) genes illustrating epigenome editing by the CRISPRi technology, as measured by CUT&RUN sequencing (H3K4me3, H3K9me3) relative to non-targeting control. I, J Reverse transcription quantitative PCR (RT-qPCR) and Western blot analysis of MORC2 (left) and TASOR (right) transcript and protein levels in MORC2 and TASOR CRISPRi cells, relative to a non-targeting control, 10 days post-transduction. Experiments were repeated at least once with similar results. Data points shown for the RT-qPCR are two independent replicates, each constituting the average of three technical replicates. Source data are provided in a Source Data file.

Fig. 2. Young FL-L1s are silenced by DNA methylation but not HUSH-MORC2 in hNPCs.

A Genetic structure of a full-length (>6-kb) intact L1 retrotransposon. TSD, target site duplication; UTR, untranslated region; ORF, open reading frame. B Nanopore DNA methylation analysis over hg38 reference FL-L1s in control (n = 1) and MORC2 CRISPRi hNPCs (n = 1). The violin plots represent the median CpG methylation of the given set of genomic intervals and span the interquartile range. C Schematic of L1 analysis of MORC2 and TASOR CRISPRi and DNMT1-KO hNPCs. D Log₂-fold-change (LFC) of young L1 subfamilies measured by RNA-seq in MORC2 or TASOR CRISPRi hNPCs (n = 7) versus controls (n = 4) using the TEtranscripts software. DNMT1-KO RNA-seq data in hNPCs (n = 3 in control group; n = 3 in treatment group) taken from Jönsson et al.. In each case the error bars represent +/- LFC standard error calculated by DESeq2 taking all samples into account. E, F Heatmaps illustrating CUT&RUN signal enrichment of H3K4me3 and H3K9me3 in control, MORC2 CRISPRi, TASOR CRISPRi and DNMT1-KO hNPCs, plotted over young full-length L1PA families sorted by evolutionary age (top to bottom). Only signal with mapQ score >10 was used to generate signal matrices. In all cases experiments were performed 10 days post-transduction and the experiments repeated at least once with similar results. G Western blotting for L1 orf1p expression in control, MORC2 CRISPRi, TASOR CRISPRi and DNMT1-KO hNPCs. The experiment was repeated once with similar results. H Mean signal of Nanopore long-read cDNA sequencing in control (n =1), MORC2 CRISPRi (n =1), TASOR CRISPRi (n =1) and DNMT1-KO (n =1) hNPCs aligning to young full-length L1s. Only uniquely-mapping reads were retained. I Two genome browser examples of FL-L1s upregulated by loss of DNMT1 but not MORC2 or TASOR. In all tracks only uniquely-mapping reads were retained. J Differential expression analysis of selected interferon (IFN) stimulated genes in the TASOR and MORC2 CRISPRi treatments (n =7) versus control hNPCs (n =4) and in DNMT1-KO hNPCs (n =3) versus control (n =3). Shown are LFC values ±LFC standard error calculated by DESeq2, taking all samples into account. The DNMT1-KO RNA-seq data was taken from Jönsson et al.. Source data are provided in a Source Data file.

Fig. 3. DNA methylation controls the activity of HUSH-MORC2 at FL-L1s.

A Three examples of FL-L1s weakly upregulated by MORC2 and TASOR CRISPRi showing promoter H3K4me3 in control hNPCs and an intact YY1 binding site (upper). Nanopore sequencing reads suggest incomplete DNA methylation over the promoter in control hNPCs (lower). In the Nanopore reads CpGs in filled black circles are called as methylated; open blue circles are unmethylated. B Measurement of MORC2 L1 binding upon genome demethylation by DNMT1 CRISPRi. C Summary plots illustrating CUT&RUN signal enrichment of MORC2 in control and DNMT1 CRISPRi hNPCs, plotted over young full-length L1 subfamilies sorted by evolutionary age. Only uniquely-mapped reads were used to generate signal matrices. Displayed are the genomic regions spanning ±5-kb from the 6-kb L1 element. The dip over the body of the elements is due to lower average unique mappability of these youngest elements relative to the surrounding genome.

Fig. 4. Genome demethylation sensitizes young L1s to restriction by MORC2.

A Strategy for simultaneous MORC2 and DNMT1 CRISPRi. B Box plots showing normalized read counts of bulk RNA-seq experiments in control (n = 4), MORC2 CRISPRi (n = 3), control’ (n = 4), DNMT1 CRISPRi (n = 4), control” (n = 4), and MORC2/DNMT1 double CRISPRi (n = 4) hNPCs. Central bands denote medians. Boxes represent the interquartile range and whiskers represent maxima and minima. C Log₂-fold-change (LFC) of young L1 subfamilies measured by RNA-seq in MORC2 or TASOR CRISPRi hNPCs (n = 7, blue), DNMT1 CRISPRi (n = 4, green), and MORC2/DNMT1 double CRISPRi (n =4, orange) versus controls (n = 4 for each independent experiment) using TEtranscripts. In each case the error bars represent ±LFC standard error calculated by DESeq2 taking all samples into account. D Mean plots of uniquely-mapped RNA-seq read counts for different treatments versus their respective controls over reference FL-L1s. Red points indicate that the element satisfied both log₂-fold-change >2 and padj < 0.05 cutoffs as calculated by DESeq2. E, F Heatmaps of RNA-seq and H3K4me3 CUT&RUN signal plotted over young full-length L1PA families sorted by evolutionary age (top to bottom) in the given hNPC treatments. In all cases only uniquely-mapped reads were retained. G Examples of young FL-L1 transcriptional and epigenomic changes upon DNMT1 and MORC2/DNMT1 depletion. H Schematic model for regulation of young FL-L1s by DNA methylation and MORC2.

Fig. 5. HUSH-MORC2 and DNA methylation control ALRs and protocadherins.

A Mean plots of RNA-seq data in Control (n =4) and MORC2 and TASOR CRISPRi hNPCs (n = 7) illustrating differential repeat expression based on TEtranscripts. ALR is the only subfamily with |log₂ fold-change| >2 and padj <0.05. B Average CpG methylation in reads mapping uniquely to pericentromeric regions relative to the genome average (10-kb bins), illustrating relative ALR hypomethylation in this hNPC line (n = 1). The violin plots represent the median CpG methylation of the given set of genomic intervals and span the interquartile range. A two-sided Wilcoxon rank sum test with continuity correction was used to compare the methylation levels between the two sets of regions. C Example of DNA methylation status in ALR-rich pericentromeric region on chr19. In the Nanopore reads CpGs in filled black circles are called as methylated; open blue circles are unmethylated. D Log₂-fold-change (LFC) of ALRs measured by RNA-seq in MORC2 or TASOR CRISPRi hNPCs (n = 7, blue), DNMT1 CRISPRi (n = 4, green) and MORC2/DNMT1 double CRISPRi (n = 4, orange) versus controls (n = 4 for each independent experiment). In each case the error bars represent ±LFC standard error calculated by DESeq2 taking all samples into account. E Gene ontology analysis of 373 differentially-expressed genes in MORC2 and TASOR CRISPRi hNPCs (no LFC cutoff, DESeq2 padj < 0.05). F Genome browser snapshot of the clustered protocadherin locus on chr5 illustrating transcriptional and chromatin regulation by HUSH-MORC2, with a zoom-in of the TAF7 gene and PCDH gamma cluster. G Heatmaps of average log₂-transformed counts over clustered PCDH genes in the given treatments.