DNMT1-interacting RNAs block gene-specific DNA methylation

Abstract

DNA methylation was first described almost a century ago; however, the rules governing its establishment and maintenance remain elusive. Here we present data demonstrating that active transcription regulates levels of genomic methylation. We identify a novel RNA arising from the CEBPA gene locus that is critical in regulating the local DNA methylation profile. This RNA binds to DNMT1 and prevents CEBPA gene locus methylation. Deep sequencing of transcripts associated with DNMT1 combined with genome-scale methylation and expression profiling extend the generality of this finding to numerous gene loci. Collectively, these results delineate the nature of DNMT1-RNA interactions and suggest strategies for gene-selective demethylation of therapeutic targets in human diseases.

Figure 1. Characterization of the ecCEBPA

a, Diagram of CEBPA transcripts; b, Assessment of transcripts by Northern blot hybridization. c–e, Relative levels of the transcripts in cellular fractions. In panel d, ecCEBPA levels are shown on different scales. qRT-PCR, bars indicate mean ± s.d. (n=3).

Figure 2. Loss- and gain-of-function studies demonstrate that ecCEBPA maintains CEBPA expression by regulating methylation of the CEBPA locus

a, Diagram indicates: position of target sequences for shRNA constructs (sh1–3); the fragment derived from the ecCEBPA employed for overexpression (R1) regions analyzed for changes in DNA methylation (distal promoter; coding sequence, CDS; and 3′UTR); b–c, The results of ecCEBPA loss-of-function in CEBPA-expressing U937 cells. Effect of ecCEBPA-targeting shRNAs on CEBPA mRNA levels. qRT-PCR, bars indicate mean ± s.d. (b) and methylation of the CEBPA promoter (c). DNA methylation changes are shown as the ratios of methylated to unmethylated CpGs in all clones analyzed per each construct (n=14); d–e, The results of ecCEBPA gain-of-function studies in K562 cells, in which CEBPA is methylated and silenced. d, Effect of ecCEBPA upregulation on CEBPA mRNA levels. UR = unrelated region. qRT-PCR, bars indicate mean ± s.d. (n=4); e, Effect of ecCEBPA upregulation on methylation of the CEBPA locus (DNA methylation changes were assessed as described in c; (n=14, for distal promoter; and n=6, for CDS, 3′UTR); f–g, The results of transcription inhibition in U937 cells. f, ecCEBPA expression levels after treatment with Actinomycin D and ML-60218 in synchronized and unsynchronized cells. qRT-PCR, bars indicate mean ± s.d.; g, DNA methylation changes after treatment with Actinomycin D and ML-60218 in synchronized (n=12) and unsynchronized (n=10) cells (assessed as described in c). All bisulfite sequenced clones were analyzed by Fisher’s exact test. *P<0.05; **P<0.01; ***P<0.001.

Figure 3. ecCEBPA–DNMT1 interactions; DNMT1 binds to RNA with greater affinity than to DNA

a, Diagram showing position of qRT-PCR primers used in RIP, double-headed arrow; RNA and DNA oligonucleotides used in EMSA and REMSA. Asterisks indicate position of methylated cytosines; umDNA, hmDNA, and mDNA refer to unmethylated, hemimethylated, and methylated DNA probes, respectively; b, ecCEBPA is immunoprecipitated with anti-DNMT1 antibody. qRT-PCR, bars indicate mean ± s.d.; c, RNA- DNMT1 binding is not affected by the absence of CpG dinucleotides (right panel). Left and middle panels: RNA oligonucleotide R2 and its mutated form mut R2 (asterisks indicate cytosines substituted into uridines), both able to form stem-loop-structures; d, RNA oligonucleotide able to form stem-loop structure bind DNMT1 (R6); e, R5 RNA oligonucleotide forming stem-loop structure (R5) has a greater DNMT1 affinity compared to mut R5, unable to fold into stem-loop, (taken in equimolar amounts), at the same DNMT1 concentration; f, Left four panels: REMSA and EMSA performed with the fixed concentration of ssRNA and dsDNA oligonucleotides (1 nM) and increasing concentrations of DNMT1 protein; Right panel: Nonlinear regression analysis of bound RNA/DNA versus DNMT1 concentrations. Error bars indicate s.d. from two independent experiments; g, REMSA showing that RNA oligonucleotide R4, which is unable to form stem-loop structure, displays lower DNMT1 affinity as compared to R5 (Fig. 3f left panel) at the same DNMT1 concentrations; h, Left panel: Schematic diagram showing the DNMT1 domains and the GST-DNMT1 isolated fragments (F1–F5); Right panel: GST-DNMT1 pull down assay demonstrating binding of the folded RNA oligonucleotide R5 to the catalytic domain of DNMT1.

Figure 4. Transcription impedes DNA methylation

a–d, Diagram showing the parallel in vitro transcription-methylation assays performed on a hemimethylated template containing the T7 promoter (Supplementary Fig. 4) with and without combinations of RNA polymerase, DNMT1, or both; e, DNMT1 exerts enzymatic activity only in the absence of transcription. COBRA analysis of methylation patterns acquired in reactions shown in b–d; f, DNA methylation changes as are shown as the ratios of methylated to unmethylated CpGs in all clones analyzed per each construct (n=5). The same effect was observed with two different RNA polymerases: T7 and Sigma-Saturated (σ70)-Holoenzyme (E. coli RNA Polymerase). DNA methylation changes were analyzed by Fisher’s exact test (*P<0.05; **P<0.01; ***P<0.001); g, in vitro DNMT1 assay demonstrating DNMT1 enzymatic impairment by RNA oligonucleotides. The assay was performed using ecCEBPA-related and unrelated RNA oligonucleotides. Sequences and position of the ribooligonucleotides are shown on Fig. 3a and Supplementary Fig. 3. Error bars indicate mean ± s.d. (n=2).

Figure 5. Genome-wide alignment of DNMT1 bound and unbound transcripts, DNA methylation, and gene expression

a, Two-way Venn diagram showing DNMT1 specific peaks overlapping with transcribed elements identified in HL-60 total and poly(A+)-depleted RNA-Seq libraries. b, Cloud plots representing genes within DNMT1 unbound, bound and all RRBS-covered groups stratified by DNA methylation and expression levels. All genes are presented in Supplementary Excel File 2. c, Examples of genes from the C (CEBPA) and B (USP29) clusters. Peaks are visualized using the SSIRs. d, Model of DNMT1 sequestration. Upper panel: DNMT1 can access transcriptionally inactive hemimethylated genomic regions. Lower panel: DNMT1 cannot access transcriptionally active hemimethylated genomic regions.