Missense mutations in the PWWP domain of DNMT3A can cause cytoplasmic mislocalization of DNMT3A in AML cells

Abstract DNMT3A is a de novo DNA methyltransferase that is highly expressed in hematopoietic stem/progenitor cells. Mutations in this gene are the most common cause of clonal hematopoiesis and are among the most common initiating events for adult de novo acute myeloid leukemia (AML) patients. DNMT3A contains three well annotated domains: the PWWP domain, the ADD domain, and the methyltransferase domain. Hundreds of AML-associated missense mutations in all three domains have been identified, and all appear to cause loss-of-function. Several mechanisms contributing to loss-of-function have been described (e.g. reduced protein stability, reduced methyltransferase enzyme activity, etc.), but the consequences of most missense mutations remain unclear. The methyltransferase activity of DNMT3A requires its ability to form homodimers or tetramers. Methyltransferase activity is also augmented by interactions with DNMT3L, or an inactive splice isoform of DNMT3B (DNMT3B3). Because these interactions are critical for DNMT3A function, we hypothesized that altered protein:protein interactions may be relevant for how some missense mutations reduce DNMT3A function. To address this hypothesis, we fused (in frame) a full length DNMT3A1 cDNA to a promiscuous biotin ligase cDNA (TurboID) in an MSCV-based retroviral vector with an IRES-GFP tag to allow for purification of transduced cells. The DNMT3A1-TurboID (TID) “bait” protein labels nearby proteins (within 10Å of the bait). We introduced 20 different, recurrent DNMT3A missense mutations associated with AML or DNMT3A overgrowth syndrome, and expressed all mutants in lineage negative, wild type (WT) mouse hematopoietic progenitors, purified GFP+ cells two days after transduction, and treated the intact cells with biotin for 4 hours; labeled proteins were purified with streptavidin beads for mass spectrometry. A total of 72 independent samples were analyzed (10 with TurboID alone, 9 with WT DNMT3A1, and 2-4 for each mutation). WT DNMT3A-TID, and 17 of the 20 mutant forms of DNMT3A1, interacted predominantly with nuclear proteins and nuclear complexes; notably, 13 of the 14 annotated members of the NuRD complex (Nuclear Remodeling and Deacetylase) interacted significantly with DNMT3A1, a previously unreported interaction. Surprisingly, 3 mutations found in the PWWP domain of DNMT3A (G298W, I310N, and R326C) predominantly interacted with cytoplasmic proteins; these mutations did not alter known nuclear import signals, or create novel nuclear export signals. Remarkably, a mutation at R301W did not cause cytoplasmic mislocalization. We validated mislocalization by transfecting these mutant isoforms into 293T cells and confirmed their cytoplasmic localization by immunofluorescence. Importantly, immunofluorescence imaging performed on primary human AML samples with DNMT3A mutations at G298E and L344R confirmed that most DNMT3A in these AML cells was cytoplasmic.We are now performing experiments designed to define the mechanism(s) underlying cytoplasmic mislocalization (e.g. loss of a nuclear protein interaction that retains DNMT3A in the nucleus, vs. gain of an interaction that sequesters DNMT3A in the cytoplasm). We therefore expressed one of the cytoplasmic mutants (DNMT3A1G298W-TurboID) in hematopoietic progenitors, and labeled with biotin either before or after cell lysis. Cells lysed prior to biotin labeling would allow TID to label proteins in any compartment of the cell, while those treated with biotin in intact cells would have labeling restricted to the compartment in which the protein normally resides. The G298W mutation eliminated interactions with one of the NuRD component proteins (SIN3A) with both conditions, suggesting that this mutation directly alters the binding of DNMT3A1 with SIN3A. Modeling of the interaction of the PWWP domain of DNMT3A1 with SIN3A (using the AlphaFold3 algorithm) revealed a high confidence interaction with protein interfacial contacts between the SIN3A PAH-3 domain, and DNMT3A residues 295, 297, 305, 306, 309, and 331. Remarkably, the R301W mutation (which did not cause cytoplasmic mislocalization) was not part of this interface. Taken together, these data suggest that DNMT3A1, via direct interactions with at least one NuRD complex protein (SIN3A), is retained in the nucleus because of this interaction, suggesting its importance for DNMT3A function. Similar experiments with several other mutations in the PWWP domain are in progress.