Background Myelodysplastic syndromes (MDS) are a heterogeneous group of diseases characterized by ineffective hematopoiesis. Hypomethylating agents (HMA) such as decitabine and azacytidine (AZA) induce clinical responses in both lower-risk (LR) and higher-risk groups; however, the treatment eventually fails and can be followed by leukemic transformation. The mechanism underlying transient clinical benefit from low-dose HMA (restoration of hematopoiesis) is not fully understood. Here we used cutting-edge single-cell multi-omic approaches to describe the abnormal clonal architecture and transcriptome in LR-MDS and how they are altered following HMA therapy. Results First, we defined genomic differences between aged and young healthy hematopoietic stem and progenitor cells (HSCPs) by building a single-cell atlas of human HSCP with molecularly-defined concentrations of relevant CITE seq antibodies. 350k high-quality bone marrow nucleated cells were obtained from four young donors; varying race and gender. A 6-fold molecular titration of 278 CITE-seq antibodies determined those which are functional and which contribute to clustering. Next, we repeated the captures on samples from aged healthy people (n=3), MDS patients (n=4) and conducted differential analysis on cell states and transcriptomes. Compared to young marrow, healthy marrow from elderly donors revealed global upregulation of inflammatory signals and innate immunity. In erythroid progenitors (ERP) we identified abnormally high expression of the transcription factors KLF1 and GATA1, but lower than normal expression of their hemoglobin target genes. At the same time altered erythropoiesis in the elderly appeared to be compensated by increased frequency of multilineage progenitors and more mature erythroblasts mitigating clinical anemia. Comparing MDS to aged marrow, while most cell populations from MDS showed significant gene expression changes, the amplitude was dwarfed by abnormal mRNA processing gene expression in ERP and erythroblasts. Moreover, in de novo LR-MDS samples and an anemia sample (without known mutations) further downregulation of hemoglobin genes in the erythroid lineage was accompanied by decreased erythroblast numbers leading to anemia. To study the impact of AZA on LR-MDS, we longitudinally profiled a single MDS patient (pre-HMA, post-HMA, sAML transformation, cycle 2 VEN+AZA) (Panel 1). We found that 6 cycles of AZA treatment depleted HSC; whereas lymphoid-primed multipotential progenitors (LMPP), megakaryocyte-erythroid progenitors (MEP), and less mature erythroblasts were enriched after AZA therapy (Panel 2 left). Using genotyping of the transcriptome (GoT) (Panel 2 right), we defined cells with mutations to find that not only SRSF2-relevant pathways (eg. mRNA processing, and mRNA splicing) are molecularly corrected (downregulated) but also there are fewer mutant cells among primitive stem cells and lineage-committed progenitors. However, mutant ERP and erythroblasts accumulated and pathways were more active suggesting the blockade of differentiation at terminal erythropoiesis is not rescued, consistent with the clinical observation that the patient was still RBC transfusion-dependent despite platelet recovery. Notably, these mutant erythroid cells, while defective, never transformed to leukemia. The expansion of mutant multilineage progenitors and LMPP preceded AML transformation and were eliminated (as evidenced by flow and CITE-seq) following two cycles of VEN+AZA. Conclusions: Age-matched references are critical in differential analysis of hematological diseases like AML or MDS. With matched elderly donor data sets, we were able to molecularly isolate functional disruption of erythropoiesis and megakaryopoiesis underlying anemia and thrombocytopenia, respectively, observed in an LR-MDS patient. AZA rescued the dysregulated signals in SRSF2-mutated LR-MDS patient HSPCs, but not in ERP and erythroblasts, and therefore only partially restored hematopoiesis. However, as AZA drove differentiation of both wildtype and mutant HSCP, accumulation of mutant multilineage progenitors and LMPPs contributed to progression to AML. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal