DNA methylation mediated by the de novo DNA methyltransferases (DNMT3A and DNMT3B) is essential for normal embryonic development. Somatic loss-of-function or mutations in DNMT3A are common in individuals with clonal hematopoiesis, and are found in nearly 30% of normal karyotype AML patients; mutations at amino acid R882 represent a hotspot in AML, accounting for ~60% of DNMT3A mutations in AML. Acquired germline DNMT3A mutations are associated with the DNMT3A Overgrowth Syndrome (aka Tatton-Brown-Rahman Syndrome/TBRS), a leukemia susceptibility syndrome. Mutations at R882 have a dominant negative effect on WT DNMT3A, and cause a distinctive, focal, canonical DNA hypomethylation phenotype. Many non-R882 DNMT3A missense mutations also have a canonical hypomethylation phenotype in hematopoietic cells that is similar to that of haploinsufficiency caused by gene deletion (Smith et al, Nat. Comms, 2021), suggesting that they are also loss-of-function mutations. Although some non-R882 missensemutations may cause DNMT3A protein instability to cause loss-of-function (Huang et al, Can Disc 2022), others do not yet have a mechanism described. For this study, we investigated the hypothesis that missense mutations in DNMT3A may inactivate its methyltransferase activity by altering its ability to interact with proteins that influence its function. To better understand mechanisms involving altered protein interactions, we used a proximity labeling system using a highly active biotin ligase (TurboID/"TID"), fused in-frame to full length human DNMT3A1 at either the N-or C-terminus in an MSCV-based vector tagged with GFP. Vectors with either WT or mutant DNMT3A cDNAs were transduced into murine bone marrow progenitors, which were allowed to grow in culture for 2 days, enriched for GFP+ cells, and then labelled with biotin for 4 hours. Proteins within 10 angstroms of the TurboID-DNMT3A fusions were tagged with biotin, enriched with a streptavidin bead pulldown, and identified using mass spectrometry. We evaluated protein interactions with TurboID alone (n=9 independent experiments), wildtype DNMT3A (n=9), or with 21 different, recurrent missense mutations (7 TBRS mutations, 4 AML mutations, and 10 mutations found in both AML and TBRS patients, n=2-4 per mutation, with 53 total, independent samples) in mouse hematopoietic stem/progenitor cells (HSPCs). ETV3, RUNX1, DNMT3B, MBD1, GATA1, KDM5A, RB1, and 428 other proteins interacted specifically with wild type DNMT3A, compared to TurboID alone. Although some of these proteins had previously been identified as DNMT3A interactors, the vast majority were previously unknown (see Fig. 1A for the 50 strongest interactors). Altered DNMT3A protein interactions caused by missense mutations are shown for 10 selected proteins in Fig. 1B. Three mutations located in the PWWP domain of DNMT3A (G298W, I310N, and R326C) caused a loss of nearly all interactions, which was found to be due to mislocalization of DNMT3A to the cytoplasm by these variants (data not shown). Mutations known to cause DNMT3A instability (Huang et al, Can Disc 2022) including C583Y, R635W, Y660H, R736H, R749C, F902S and P904L, had increased interactions with ubiquitin ligases (Hectd3, Rnf31, Ube2v1, Smurf1, Trim8) which may target them for accelerated degradation (data not shown). The normal interaction between DNMT3A and DNMT3B was essentially eliminated by several mutations (I310N, R635W, Y660H, R736H) and decreased in several more (G298W, F902S and P904L) (Fig.1B). In AML cells, the dominant isoform of DNMT3B is DNMT3B3, which is missing exons that are essential for its methyltransferase activity. However, DNMT3B3 is known to interact with DNMT3A and, in doing so, increase its activity by acting as a chaperone. Therefore, the disruption of DNMT3B3 binding to DNMT3A may represent a novel mechanism for missense mutations to reduce DNMT3A activity. Finally, we used the same system to define DNMT3A protein interactions in human CD34+ cord blood cells (with DNMT3A WT, R736H, and R882H, and others in progress), which revealed high concordance with the mouse data. DNMT3A missense mutations can alter many protein interactions with DNMT3A, which can clearly reduce its methyltransferase activity by a number of mechanisms. These findings may suggest new approaches for restoring DNMT3A activity in patients with myeloid malignances. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal
Read full abstract