Base Editor Scanning Reveals Activating Mutations of DNMT3A

Abstract

DNA methyltransferase 3A (DNMT3A) is a de novo cytosine methyltransferase responsible for establishing proper DNA methylation during mammalian development. Loss-of-function (LOF) mutations to DNMT3A, including the hotspot mutation R882H, frequently occur in developmental growth disorders and hematological diseases, including clonal hematopoiesis and acute myeloid leukemia. Accordingly, identifying mechanisms that activate DNMT3A is of both fundamental and therapeutic interest. Here, we applied a base editor mutational scanning strategy with an improved DNA methylation reporter to systematically identify DNMT3A activating mutations in cells. By integrating an optimized cellular recruitment strategy with paired isogenic cell lines with or without the LOF hotspot R882H mutation, we identify and validate three distinct hyperactivating mutations within or interacting with the regulatory ADD domain of DNMT3A, nominating these regions as potential functional target sites for pharmacological intervention. Notably, these mutations are still activating in the context of a heterozygous R882H mutation. Altogether, we showcase the utility of base editor scanning for discovering functional regions of target proteins.

Figure 1:. Strategy for and validation of DNMT3A activating screen

a) Cartoon showing the original base editor scanning and updated strategy for identifying DNMT3A activating mutations. By lowering the background activity of our reporter, we decreased the probability that activating mutations would be hidden by the limited dynamic range between baseline and maximum DNMT3A activity. dDNMT3A = codon-switched DNMT3A2^E756A. b) Cartoon showing our strategy for base editor scanning and separating cells based on DNMT3A activity. c) Line plot showing silencing over time for rTetR-DNMT3L (v1), DNMT3A^WT or DNMT3A^WT/R882H (v2) cells. See Figure S2c. for significance testing. d) Histogram showing silencing of citrine reporter in cells with DNMT3A^WT or DNMT3A^WT/R882H after 12 days of dox treatment. See Figure S2d. for significance testing. e) Line plot showing silencing over time of DNMT3A^WT or DNMT3A^WT/R882H cells with and without the base editing library. Data are mean ± SD of n = 3 replicates. See Figure S2e. for significance testing.

Figure 2:. Results of base editor screens in paired isogenic cell lines

a) Scatter plot of sgRNA enrichment across the coding sequence of DNMT3A in DNMT3A^WT cells. sgRNA score corresponds to log₂ fold-change of citrine-negative cells/unsorted cells. Enrichment scores for all guides can be found in Table S4. b) Same as a) but in DNMT3A^WT/R882H cells. Enrichment scores can be found in Table S5. c) Box plot showing enrichment of sgRNAs sorted by category in both screens. sgRNAs in the gray box are predicted to edit within a DNMT3A exon, while those in other categories either edit outside of DNMT3A exons or are not predicted to edit. d) Pymol cartoon model showing predicted mutated residues from missense activating (left) and inactivating (right) sgRNAs mapped to the crystal structure of the DNMT3A ADD and MTase domains (PDB: 4u7p).

Figure 3:. Validation of sgRNA hits

a) Scatter plot showing sgRNA scores in DNMT3A^WT vs DNMT3A^WT/R882H reporter cells. sgRNAs tested in b) and c) are highlighted. Spearman R = 0.51, p=8.1×10⁻⁴⁴ b) Bar plots showing % silenced cells treated with single sgRNAs after 10 (DNMT3A^WT) or 17 (DNMT3A^WT/R882H) days of dox treatment. sgLuc = Luciferase-targeting sgRNA, NIC = non-infection control. Data are mean ± SD of n = 3 replicates. * Correspond to significance testing using two-sided students t-tests comparing each guide to sgLuc with Bonferroni multiple hypothesis correction. *p ≤ 0.0167, **p ≤ 0.0033, ***p ≤ 0.00033, ****p ≤ 3.33 ×10⁻⁵ See Figure S4 for additional sgRNAs and replicates and table S6 for a summary of all sgRNAs tested. c) PyMOL models showing the structural location of key activating mutants. PDB IDs: 4u7t (active) and 4u7p (autoinhibited).

Figure 4:. Genotyping of sgRNA hits

a) Tables showing base editing outcomes with >5% allele frequency from treatment of DNMT3A^WT reporter cells with single missense sgRNAs (See Figure S5 for DNMT3A^WT/R882H cell line). Analysis performed with CRISPResso2. b) Scheme showing genomic DNA and cDNA sequencing of DNMT3A^WT reporter cells treated with sgIntron-1. (See Figure S5 for DNMT3A^WT/R882H cell line). c) PyMOL model showing location of D618_K621del (PDB ID 4u7p).

Figure 5:. Biochemical validation of activating mutants

a) Bar plot showing fold-change in DNMT3A methyltransferase activity of mutants vs WT, based on data in b). P-values were calculated by unpaired two-tailed students’ t-tests. Fold-change errors were propagated from the individual standard deviations. b) Bar plot showing methyltransferase activity of DNMT3A mutants ± 5 μM H3K4me0. Data are mean ± SD of n = 3 replicates. P-values were calculated by unpaired two-tailed students’ t-tests. See Figure S7 for SDS-PAGE gel of purified proteins and replicate 2.