DNMT3A R882H Exhibits Greater Inflammatory Potential Than R882C in Primary Hematopoietic Stem and Progenitor Cell Knock-in Model and Population Data

Clonal hematopoiesis (CH) is an age-associated phenomenon which is known to increase the risk for hematologic malignancy and cardiovascular disease. The most commonly mutated gene in CH is de novo DNA methyltransferase DNMT3A. R882H and R882C are the most common hotspot mutations in this gene, and these mutations are commonly grouped together in interpretation of clinical data even though in vitro experiments suggest divergence in their biochemical consequences. Here, we used CRISPR knock-in of primary hematopoietic stem and progenitor cells (HSPCs) along with population data from the UK Biobank (UKB) to evaluate the extent to which R882H and R882C differentially predispose to inflammatory phenotypes.

Using two homology donor (HD) templates with distinct self-cleaving fluorescent tags, we used FACS to isolate homozygous R882H- or R882C-mutant cells from six unique samples of cord blood HSPCs. We subsequently cultured these isogeneic mutant cells and their mock-electroporated wild-type (WT) controls in monocyte differentiation media and evaluated their transcriptomes via bulk RNA sequencing (RNAseq) and production of secreted cytokines in a Luminex-based assay.

Analysis of transcripts using differential expression for repeated measures (Dream) revealed that, when compared to WT, the mutants demonstrated many shared differentially expressed genes (DEGs) but also exhibited many DEGs unique to each variant. Furthermore, gene set enrichment analysis (GSEA) of Hallmark pathways revealed common enrichment of cell cycle related pathways in both variants, but found that R882H displayed greater enrichment of inflammation-related pathways than either WT or R882C (Figure 1A). Rank-order GO and KEGG pathway enrichment comparison of R882H vs R882C showed further evidence of bias in inflammation-related pathways among the most upregulated gene sets (e.g., GO:0019882, Antigen Processing & Presentation and KEGG:hsa04145, Phagosome). IL-6 is significantly regulated by R882 genotype in our model; levels of IL-6 for R882H (13,065 ± 2,446; mean ± s.e. in pg/mL) but not R882C (10,340 ± 2,446) were significantly different than WT (6,281 ± 1,730; p = 0.014).

We next sought to better understand if this biased inflammatory priming would also be observed for heterozygous mutations and across different cell types, so using mutant and WT HD templates we performed single cell RNAseq on CRISPR’d heterozygous mutants and knock-in WT controls. Comparing R882H to R882C in this context demonstrated upregulation of numerous inflammation-related genes, including CXCL8 (IL-8), S100A8, and S100A9. GSEA of Hallmark pathways revealed enrichment of inflammatory pathways (e.g., Inflammatory Response & TNFA Signaling via NFKB) in monocytic cells but also in progenitor cells.

We then investigated whether these transcriptional differences between R882H and R882C might be due to divergent patterns of DNA (hypo)methylation. Employing 5-letter next-generation sequencing (A, C, G, T, methyl-C), we compared the methylation landscape of three isogeneic donors with homozygous R882H, R882C, or mock-electroporated WT genotypes. As expected, comparison of either mutant to WT showed a global hypomethylation. Comparison of R882H to R882C revealed no systematic bias in methylation of gene promoters but a pervasive hypomethylation of gene bodies. A gene set of the 50 most upregulated genes from our single cell RNAseq experiment was significantly enriched in rank order analysis of R882H vs R882C hypomethylation of gene bodies (NES = 1.47, p = 0.036) but not promoters (NES = -0.86, p = 0.71).

Finally, we asked if there were differences in patient phenotypes for those with R882H or R882C clonal hematopoiesis. Here, we used our previously published UKB clonal hematopoiesis cohort (Vlasschaert et al. 2023, Blood). In Cox proportional hazards models controlling for basic demographics (N = 451K) or demographics plus cardiovascular-relevant covariates (N = 390K), we found that R882H but not R882C was associated with significantly greater incidence of heart failure and of a composite measure consisting of death or coronary artery disease (Figure 1B). In conclusion, our study found evidence in experimental models and patient data that DNMT3A R882H may pose a greater risk for inflammatory phenotypes than R882C.