A novel requirement for DROSHA in maintenance of mammalian CG methylation

Abstract

In mammals, faithful inheritance of genomic methylation patterns ensures proper gene regulation and cell behaviour, impacting normal development and fertility. Following establishment, genomic methylation patterns are transmitted through S-phase by the maintenance methyltransferase Dnmt1. Using a protein interaction screen, we identify Microprocessor component DROSHA as a novel DNMT1-interactor. Drosha-deficient embryonic stem (ES) cells display genomic hypomethylation that is not accounted for by changes in the levels of DNMT proteins. DNMT1-mediated methyltransferase activity is also reduced in these cells. We identify two transcripts that are specifically upregulated in Drosha- but not Dicer-deficient ES cells. Regions within these transcripts predicted to form stem-loop structures are processed by Microprocessor and can inhibit DNMT1-mediated methylation in vitro. Our results highlight DROSHA as a novel regulator of mammalian DNA methylation and we propose that DROSHA-mediated processing of RNA is necessary to ensure full DNMT1 activity. This adds to the DROSHA repertoire of non-miRNA dependent functions as well as implicating RNA in regulating DNMT1 activity and correct levels of genomic methylation.

Figure 1.

Drosha interacts and co-localizes with Dnmt1. (A) Western blot using anti-Drosha antibody following immunoprecipitation with anti-FLAG M2 magnetic beads on material from wild-type (+/+) and Dnmt1^Tag/+ (Tag/+) ES cells. (B) Immunofluorescence micrographs of COS-7 monkey kidney cells co-transfected with DsRed-DNMT1 and HA-Drosha constructs. (C) Cartoon showing different fragments of Drosha examined and the domains they encompass. (D) Scatter plot showing distribution of cells with different Manders Split Coefficient values corresponding to co-localization between DsRed-DNMT1 and the different HA-Drosha constructs indicated. ‘–’ indicates cells transfected with DsRed-DNMT1 only. Dnmt1 N terminal fragment contains amino acids 2–452. Exact Mann–Whitney P-values of different fragments compared to distribution using full-length HA-Drosha are Fr1 (0.0201), Fr2 (0.2872), Fr3 (0.3934), Fr4 (0.4197), Fr5 (0.0001), Fr6 (0.3403), DNMT1 N-term (0.1636) and Drosha E1045Q (0.211). (E) Bar graph showing percentage of cells analysed with Manders Split Coefficient values of >0.4. Images from at least 50 cells were analysed per construct.

Figure 2.

Generation and characterization of Drosha-deficient ES cells. (A) CRISPR/Cas-mediated gene targeting strategy used to inactivate endogenous Drosha locus. gRNA sequence is shown with hCas9 cut site highlighted in red. PCR primers used for screening are indicated by arrows (A,-E). Alleles generated in Drosha^−/− ES cells following CRISPR/Cas-mediated gene editing. Actual sequence of alleles is shown in Supplementary Figure S3. (B) Agarose gel electrophoresis of PCR amplicons used to diagnose targeting of Drosha locus. Primers are indicated in (A). (C) Western blot using anti-Drosha antibodies recognizing different regions of DROSHA protein on whole cell extracts from wild-type and Drosha^−/− ES cells. Anti-tubulin antibody used as a loading control. Note that the lower band observed using the N-terminal antibody is a non-specific signal. (D) Box-whisker plots of RNA-seq data from wild-type and Drosha^−/− small RNA libraries. P = 4.182 × 10⁻¹⁰ (Wilcoxon-rank sum test). Expression levels (log₂ fragments per kb of exon per million fragments (fpkm)) of 162 miRNAs. (E) Pie chart indicating percentage of miRNAs in different categories caused by Drosha-deficiency.

Figure 3.

Drosha-deficiency results in global hypomethylation. (A) Bar graphs showing LC-MS/MS results measuring percentage of 5mC in genomic DNA extracted from wild-type (+/+) and Drosha-deficient (−/−) ES cells. Exact P-value = 0.0007 (Mann–Whitney test). (B) Same as (A) but measurements made on genomic DNA extracted from wild-type and various Drosha-deficient ES cell clones. (C) Same as (A) but measurements made on genomic DNA extracted from wild-type, Drosha-deficient ES cells and two independently transduced clones of Drosha-deficient ES cells expressing Drosha cDNA (+Drosha cDNA). (D) Box-whisker plots showing the distribution of methylation levels of each cytosine in CG context represented in WGBS data generated from the cell lines indicated. (E) Notched box-whisker plots showing weighted CG methylation levels in the various compartments in the different cell lines indicated; c/c indicates Dnmt1^c/c, which are deficient for DNMT1 protein. Exact P-values (Wilcoxon-rank sum test) for comparisons between wild-type and Drosha^−/− are P < 2.2 × 10⁻¹⁶. (F) Metaplots of CG methylation within gene bodies for 15 000 randomly selected RefSeq genes, genes up- and downregulated in Drosha^−/− compared to wild-type ES cells. Deregulated genes defined as those with a >1.5-fold change and P-value < 0.05 in Drosha^−/− mRNA-Seq dataset compared to wild-type. All bar graphs show mean +/− standard deviation.

Figure 4.

Analysis of components involved in regulating DNA methylation (A). Western blot data using the various antibodies indicated on nuclear extracts from the ES cell lines indicated. Triple Knockout (TKO) is an ES cell line deficient for all three active DNA methyltransferases (Dnmt1^−/−Dnmt3A^−/−Dnmt3B^−/−). (B) Bar graph of RNA-seq data showing expression levels of genes indicated. (C) Bar graphs of qRT-PCR data showing expression levels of the factors indicated. Results representative of three biological replicates. For Uhrf1 levels, exact P-value = 0.0571 (Mann–Whitney test). For all other genes, P = 0.028. Gapdh used for normalization. (D) RNA-Seq data traces for Dnmt3A and Dnmt3B loci. (E) Bar graphs of RNA-seq data of different Dnmt3 isoforms indicated. (F) Same as (E), but showing Image-J quantified protein levels. (G) Western blot data using anti-Tet1 and Tet2 antibody. Anti-tubulin used as loading control.

Figure 5.

DROSHA upregulated RNAs are processed by Microprocessor and inhibit DNMT1-mediated methyltransferase activity. (A) Bar graph showing ³H-labelled SAM incorporation into hemi-methylated DNA by lysate from ES cells with the genotypes indicated (P = 0.022, Mann–Whitney test). (B) Volcano plot of log₂ fold-change versus log₁₀P-value of 21 509 RefSeq genes in Drosha^−/− compared to wild-type ES cells. Previously identified DNMT1 Interacting RNAs (DiRs) not upregulated in Dicer^−/− microarray data highlighted in red, as well as Mir17hg and Dgcr8. (C) Bar graphs showing qRT-PCR expression data of factors highlighted in (B). Results representative of two biological replicates. P-values (Mann–Whitney test) are: Pdia4, 0.4857 (wild-type versus Dicer^−/−); Hspa5, 0.2, >0.99, 0.3429 (wild-type versus Drosha^−-/−, wild-type versus Dicer^−-/−, Drosha^−/− versus Dicer^−/−, respectively). P = 0.028 for all other comparisons. All bar graphs show mean +/− standard deviation. Gapdh used for normalization. (D) Graphic showing predicted hairpin/stem loop regions within Cd97 and Fbxo15 RNA. Numbers indicate base positions. (E) Western blot of lysate from HEK293T cells transiently transfected with FLAG-tagged DROSHA and DGCR8 (+MP) probed with anti-FLAG antibody. –MP indicates material from untransfected cells. (F) Autoradiograph of polyacrylamide gel-resolved in vitro transcribed RNAs indicated incubated with cell extracts from untransfected (–MP) or Microprocessor transfected (+MP) cells. Note the great reduction in signal of unprocessed full-length RNA and appearance of lower molecular weight smear in +MP lanes compared to –MP lanes. (G) Bar graph of in vitro methyltransferase assays. Cells were transfected with either DNMT1 alone or DNMT1 along with the in vitro transcribed RNAs indicated.