ALBERT

Description: Higher order chromatin structure and DNA methylation are implicated in multiple developmental processes, but their relationship to cell state is unknown. In order to understand how the DNA methylation is connected with nuclear architecture and can vary between cell types and during cell differentiation, we began to explore the 3D architecture of human hematopoietic stem and progenitor cells (HSPCs) by performing in situ Hi-C experiments at 5kb resolution. We found that large (~10kb) DNA methylation canyons can form long loops connecting anchor loci that may be dozens of megabases apart. These canyons also can form interchromosomal links (Fig.1a and 1b). We further confirmed these long-range interactions by performing 3D-FISH using two color fluorescent labeled probes that spanned the HOXA locus loop anchor (green) and the SP8 locus loop anchor (red), which are ~7MB apart (Fig. 1c). In order to begin to investigate mechanisms that may regulate these long loops and how they relate to commonly studied loops that are mediated by CTCF-extrusion, we examined their properties systematically. Interestingly, the anchors of long loops exhibited minimal enrichment for CTCF (1.04-fold), and, even when CTCF was bound, they did not obey the convergent rule. The data suggest these loops are formed by phase separation of the interacting loci to form a genomic subcompartment, rather than by CTCF-mediated extrusion. Next, we sought to determine whether other features correlated with these long loops. By aligning DNA methylation profiles with the Hi-C data, we observed that anchors often corresponded to regions of very low DNA methylation, and thus sought to analyze the relationship in detail. We found that the anchor position of the long loops had lower average DNA methylation levels than standard loop anchors and very often overlapped with DNA methylation canyons. Canyons are typically decorated with either active or repressive histone marks. We considered whether a particular group of canyons was associated with the long loops. Our findings further indicate that repressed regions marked by the polycomb-mediated histone modification H3K27me3 at DNA methylation canyons generally mediate the formation of canyon loops. Next, we considered whether the long loops associated with repressive grand canyons that we had annotated in HSPCs were present in other cell types. Using Aggregate Peak Analysis (APA), a computational strategy in which the Hi-C submatrices from the vicinity of multiple putative loops are superimposed, we examined 19 human cell types and 10 murine cell types in which loop-resolution Hi-C maps are available. Interestingly, unlike previously characterized genomic subcompartments, these long-range loops are only present in stem and progenitor cells, but not in differentiated cell types, such as T cells and erythroid progenitors (Fig. 1d). Further, we identified one particular loop anchor that lay at the anchor of a long loop and contained no apparent genes ("geneless" canyon, or "GLS"). The GLS harboring this anchor is 17 kb long, lies 1.4 Mb upstream of the HOXA1 gene, and forms long loops with a 28 kb grand canyon in the HOXA region. In order to understand the role of the GLS region in hematopoietic stem cells (HSCs), we deleted the GLS in HSPCs using Cas9-mediated editing and assayed the edited cells for their ability to form colonies. Strikingly, after deleting the GLS, the number of colonies and their size was greatly reduced in edited cells compared to control experiments using either random guide RNAs or electroporation only (Fig. 1e). After ex vivo culture, the overwhelming majority of GLS-knock out HSPCs acquired the marker CD38, indicating that they were differentiating. Similarly, HOXA gene expression, an indicator of HSPC function, was greatly diminished after GLS deletion compared to control cells. These data indicate that the GLS identified in our study is functionally associated with maintenance of the HSC state. Overall, our work reveals long-range interactions between H3K27me3-marked DNA methylation canyons comprising a novel microcompartment associated with cellular identity. Figure 1. Figure 1. Disclosures No relevant conflicts of interest to declare.

Description: Mutations in three functional domains of DNA methyltransferase 3A (DNMT3A) have been found in individuals with clonal hematopoiesis of indeterminate potential (CHIP), hematological malignancies, and cohorts of patients with overgrowth syndrome (Tatton-Brown Rahman Syndrome, or TBRS). In CHIP, the majority of DNMT3A mutations (~85%) are not at the Arginine 882 residue that is most frequently seen in AML (around 58% of DNMT3A-mutant cases). While most studies to date have focused on DNMT3A mutations in the hotspot residue R882, frameshift or premature stop codon mutations resulting in haploinsufficiency of DNMT3A also have been shown to predispose toward myeloid malignancies. However, the functional impact of these non-R882 missense mutations, and the implications for prognosis, have not yet been examined. To address this question, we generated a murine model with a single amino acid deletion of tryptophan 293 (corresponding to aa 297 in human) in the PWWP domain of DNMT3A (Dnmt3aW293Del model). We found that, similar to Dnmt3a-null mice, Dnmt3aW293Del/W293Del mice do not survive beyond postnatal day 24, suggesting that Dnmt3aW293Del is a hypomorphic mutation. In addition, Dnmt3aW293Del/+ mice recapitulated many of the features of human TBRS syndrome including obesity and neurological defects. Using methylation-deficient mouse embryonic stem cells, we re-introduced doxycycline-inducible mutant DNMT3AW297Del and observed only a negligible increase in DNA methylation, while intact DNMT3AWT protein showed a global 60% increase of DNA methylation, as measured by whole genome bisulfite sequencing. To determine the molecular mechanisms through which DNMT3AW293Del acted as DNMT3A-null mutation, we first examined both RNA and protein expression in the Dnmt3aW293Del murine model. Quantitative PCR revealed that both Dnmt3aWT and Dnmt3aW293Del mRNA were expressed at normal levels in the Dnmt3aW293Del/+ mice. Unexpectedly, however, we found that DNMT3AW293Del mutant protein could not be detected. Using bicistronic vectors to measure protein stability in human embryonic kidney 293 cells (HEK293T), we demonstrated that loss of DNMT3AW297Del protein is not due to a translational defect but to impaired protein stability. Additionally, we found application of a proteasome inhibitor could rescue expression of mutant DNMT3AW297Del expression. Furthermore, using lymphoblastoid cell lines (LCLs) derived from patients, we found DNMT3A protein expression was lower in DNMT3AW297Del/+ LCLs and that exposure to a proteasome inhibitor could increase protein levels, reinforcing the concept that deletion of aa 297 reduced DNMT3A protein stability. To determine whether reduced protein stability was a common feature of DNMT3A mutations, we examined stability of 105 additional missense mutations across three functional domains frequently mutated in CHIP, hematological malignancies, and TBRS. Surprisingly, 45 out of 105 missense DNMT3A mutations (42.9%) examined had impaired protein stability, the majority of which were located in the PWWP and catalytic domains. Using variant allelic frequencies (VAFs) from a publicly available dataset of patients with non-hematological malignancies, we found that mutants with reduced protein stability were associated with significantly higher VAFs compared to mutants with unknown function. Notably, R882 mutants had significantly higher VAFs compared to the mutants with impaired protein stability. These data indicate that individuals with R882 mutations may have the highest likelihood of conversion to malignancies, with mutants with impaired stability having moderate likelihoods, and mutants in domains with unknown function may have lowest chance to convert to malignancies and best prognosis. In this study, we show that mice with a mutation in the PWWP domain of Dnmt3a recapitulate the phenotypes of human overgrowth patients. We further discovered that a large portion of missense DNMT3A mutations have impaired protein stability. This study highlights the importance of understanding how DNMT3A protein expression is regulated and suggests proteasome inhibition may be a potential treatment for patients with particular mutations in DNMT3A, found in overgrowth syndrome, CHIP, and hematological malignancies. Disclosures No relevant conflicts of interest to declare.

Description: Mutations in the epigenetic regulators DNMT3A and IDH1/2 co-occur in patients with acute myeloid leukemia and lymphoma. In this study, these 2 epigenetic mutations cooperated to induce leukemia. Leukemia-initiating cells from Dnmt3a−/− mice that express an IDH2 neomorphic mutant have a megakaryocyte-erythroid progenitor–like immunophenotype, activate a stem-cell–like gene signature, and repress differentiated progenitor genes. We observed an epigenomic dysregulation with the gain of repressive H3K9 trimethylation and loss of H3K9 acetylation in diseased mouse bone marrow hematopoietic stem and progenitor cells (HSPCs). HDAC inhibitors rapidly reversed the H3K9 methylation/acetylation imbalance in diseased mouse HSPCs while reducing the leukemia burden. In addition, using targeted metabolomic profiling for the first time in mouse leukemia models, we also showed that prostaglandin E2 is overproduced in double-mutant HSPCs, rendering them sensitive to prostaglandin synthesis inhibition. These data revealed that Dnmt3a and Idh2 mutations are synergistic events in leukemogenesis and that HSPCs carrying both mutations are sensitive to induced differentiation by the inhibition of both prostaglandin synthesis and HDAC, which may reveal new therapeutic opportunities for patients carrying IDH1/2 mutations.

Description: Tet2 catalyzes the conversion of 5-methylcytosine to 5-hydroxylmethylcytosine (5hmC), which is considered the first step in active DNA demethylation. TET2 is frequently mutated in various hematopoietic malignancies. Mutations in TET2 and DNMT3A often co-occur in T cell lymphoma patients, though they work in the same DNA methylation-hydroxymethylation pathway. Here, we explored whether TET2 and DNMT3A could cooperate and how they interact in malignant hematopoiesis using an Mx1-cre+; Dnmt3af/f; Tet2-/- (DKO) mouse model, using as controls Mx1-cre; Dnmt3af/f (Dnmt3a KO) mice, Tet2-/- (Tet2 KO) mice and WT mice. We performed whole bone marrow transplantation of each genotype in competition with WT bone marrow to compare the competiveness of mutant stem cells. We found the order of engraftment activity to be: DKO 〉 Tet2 KO 〉 Dnmt3a KO, suggesting Dnmt3a loss-of-function augments the competitive advantage of Tet2 KO cells. Consistent with our results, DKO recipients showed a threefold increase in Lin-Sca1+cKit+ (LSK) population compared to Tet2 KO recipients 6 months after transplantation. There was also an increase in short-term hematopoietic stem cells, showing Dnmt3a loss can increase the stem cell and progenitor pool in Tet2 KO background. Forty weeks after transplantation, DKO recipients developed lethal B cell ALL characterized by leukocytosis and anemia, while Tet2 KO mice did not show signs of hematopoietic malignancy one year after transplantation. We then isolated genomic DNA and RNA from Tet2 KO and DKO long-term hematopoietic stem cells (Lin- cKit+Sca1+ CD48- CD150+) and performed whole genome bisulfite sequencing and RNA-seq. We compared the transcriptome and methylome of DKO, Tet2 KO, and Dnmt3a KO with that of WT HSCs, and we identified the differentially expressed transcripts and DMRs in the DKO that overlap with those in the single mutants, as well as those unique to the DKO. At the transcriptome level, the DKO HSCs overexpressed multiple transcription factors associated with differentiated lineage - lymphoid Ikzf1, Pax5, and Ebf1, myeloid Cebpa and Cebpe, and erythroid Klf1. Specifically, overexpression of the erythroid gene signature controlled by Klf1 was observed only in DKO HSCs. This signature is also found in AML patients carrying both DNMT3A and TET2 mutations, suggesting Klf1 overexpression in DKO HSCs could be the driving event of leukemogenesis. Further, we found the expression levels of Klf1 and Ikzf1 across the genotype in following order: DKO〉Tet2 KO〉WT〉Dnmt3a KO, suggesting Tet2 represses these genes in wild type HSCs. At the methylome level, most hypomethylated differential methylated region (DMR) in DKO HSCs are already hypomethylated in Dnmt3a KO HSCs - evidence that Dnmt3a is upstream of Tet2 in the regulation of the methylome. We identified 9 major dynamic methylation patterns across the 3 genotypes, including one that indicates that Dnmt3a and Tet2 are counteractive. However, in about 15% of these DMRs, methylation levels in the DKO decreased to an average of 10%, while the same genomic regions of the Dnmt3a KO had average methylation levels of 50% (a 30% decrease compared with WT). This significant decrease in Dnmt3a KO indicates that Tet2 maintains the methylation at these DMRs in Dnmt3a KO HSCs. Further investigation shows that these DMRs are enriched in 5hmC in the Dnmt3a KO HSCs, which implies that Tet2 is hydroxymethylating the DMRs at these sites when Dnmt3a is lost. Both Klf1 and Ikzf1 harbor a DMR of this kind in their upstream promoter regions. The hypomethylated DMR associated with Klf1 is already hypomethylated in the Dnmt3a KO HSCs, but the methylation level further decreased in the DKO from 50% (Dnmt3a KO) to 0.02% (DKO). Elevated 5hmC is also observed in the DMRs of Klf1 and Ikzf1 in Dnmt3a KO HSCs. 5hmC is thought to be associated with both repressive and activating marks, and the expression of Klf1 and Ikzf1 is repressed in Dnmt3a KO HSCs but upregulated in Tet2 KO. Thus we propose that DNA methylation at regulatory regions of these factors decreases in Dnmt3a KO HSCs, but Tet2 represses their expression by hydroxylmethylation. The loss of both Tet2 and Dnmt3a is required for full activation of Klf1 and Ikzf1. Therefore, we conclude that in HSCs both Tet2 and Dnmt3a likely repress differentiation-associated transcriptional factors by hydroxymethylating and methylating at regulatory regions. Figure 1 Figure 1. Disclosures No relevant conflicts of interest to declare.