James S. Blachly

SUMMARY: Human natural killer (NK) cells arise from common innate lymphoid cell progenitor and precursor populations that are selectively enriched within secondary lymphoid tissues and that complete their terminal differentiation through stages of progressively lineage-restricted downstream developmental intermediates. We identify two subsets of tonsil CD94/NKG2A ⁺ NKp80⁻ “stage 4A” NK cell development intermediates that show differential expression of the inhibitory receptor, CD200R1. Ex vivo single-cell transcriptomic, DNA methylation, phenotypic, and functional analyses demonstrate that CD200R1⁺ stage 4A cells represent an immature, oligopotent precursor population enriched for tissue-resident and cytokine-responsive programs. In contrast, CD200R1⁻ stage 4A cells show more restricted NK cell developmental potential concomitant with features of mature NK cells. These findings help to refine the human NK cell developmental pathway in tonsils and highlight CD200R1 as a useful marker delineating an uncommitted developmental state linking microenvironmental cues to NK cell maturation and plasticity.

Gene mutations and chromosome abnormalities are important components of prognostication in acute myeloid leukemia (AML). Here we assessed whether DNA methylation patterns in AML patients can augment risk assessments provided by genetic and other markers to better predict outcomes. Unsupervised DNA methylation patterns separating patients into 13 DNA methylation subtypes (epitypes) were used to classify 1,262 patients with de novo AML. Epitypes were predominantly comprised of a predominant genetic alteration; however, some patients within epitypes lacked cardinal alterations and were termed genetic alteration-like. Interestingly, patients displaying alteration-like DNA methylation patterns of CEBPAbZIP, FLT3-ITD, core-binding factor, KMT2A-rearrangements and other abnormalities displayed outcomes similar to patients with actual cardinal alterations. We further derived a DNA methylation signature enriched in patients with FLT3-ITD mutations that involved hypomethylation of STAT binding sites, termed the STAT hypomethylation signature (SHS). SHS positivity identified patients with inferior outcomes further adding to the prognostic significance of FLT3-ITD. Machine learning modeling revealed these DNA methylation signatures together significantly added to genetic, demographic and clinical markers to predict remission, relapse and overall survival. In summary, DNA methylation signatures capture patients who mimic cardinal genetic mutations providing additional prognostic information that may be used to in concert with standard genetic markers.