Clonal Dissection of MDS and Secondary AML Resolves Shared Splicing Neoantigens and Mechanistic Underpinnings of Hypomethylating Agent Therapeutic Response

Introduction: Myelodysplastic syndromes (MDS) are incurable myeloid malignancies with genetic risk-associated progression to myelodysplasia-related (MR)-Acute Myeloid Leukemia (AML). MDS are genetically, histologically, and clinically heterogeneous hematologic disorders characterized by ineffective hematopoiesis and cytopenias. MDS is associated with mutations in hematopoietic stem cells (HSC) that most often occur during aging. Hypomethylating agents (HMA) such as decitabine and azacytidine (AZA) induce a clinical response in both low- (LR) and high-risk MDS, but only in a subset of patients. New long-read single-cell sequencing technologies have the potential to disentangle the mechanistic underpinnings, clonal relationships and splicing alterations that occur during therapy in MDS, in an unbiased manner.

Methods: To understand the mechanistic basis of HMA therapy and leukemic transformation we applied a novel multiomic single-cell approach in serial MDS bone biopsies, spanning young, aged, MDS, HMA treated and secondary AML (sAML) samples. Bone marrow HSC and progenitor cells (HSCPs) were profiled using CITE-Seq with both Illumina short-read and long-read sequencing (PacBio MAS-Seq on the same 10x Genomics 3’ single-cell libraries). Single-cell mutation and isoform predictions from MAS-Seq were validated using targeted Genotyping of Transcriptomes (GoT) and surface protein expression in the same cells. Discrete single-cell populations in healthy subjects, MDS and sAML were mapped to our recently established CITE-Seq progenitor bone marrow atlas¹. Novel age-associated stem and progenitor heterogeneity was defined using sub-clustering and cell-state differential gene expression analyses (ICGS2 and cellHarmony). Isoform, splicing and clonal impacts were determined using a new bioinformatic toolkit, AltAnalyze-LR (long-read). To define novel targets for future therapies, we identified splicing neoantigens and associated novel transmembrane isoforms using our recently published deep-learning SNAF² workflow using our MAS-Seq and prior published bulk MDS RNA-Seq cohorts.

Results: During aging, we observed profound differences in the most primitive HSPCs, early lymphoid, monocytic and dendritic progenitors, marked by significant induction of interferon pathway gene expression. These changes were accompanied by broad splicing alterations in the same cell populations, impacting hundreds of known and novel full-length mRNA isoforms. Comparison of these same single-cell populations in diagnostic MDS patient biopsies (n=11) and aged bone marrow defined a highly coordinated gene program in MDS quiescent HSC, largely associated with down-regulation of inflammatory, glucocorticoid, cytokine signaling pathway components and splicing regulators. HMA therapy in the same patients only partially rescued these malignant HSC gene programs. From the long-read single-cell data, we observe distinct clones enriched in different cells states in primary MDS, HMA and sAML samples that could be verified with targeted variant sequencing (GoT). Integration of MAS-Seq long-read isoforms with MDS bulk RNA-Seq neojunctions from 346 patients, identified hundreds of high confidence candidate splicing neoantigens, including those frequently shared among patients (>15%) and those specific to MDS patients with splicing factor mutations (e.g., SF3B1 K700E and SRSF2 P95X).

Conclusions: We demonstrate the clonal dissection of malignant hematopoiesis during MDS therapy and leukemogenesis. Age-matched references were found to be critical in determining gene expression differences in discrete HSPC cell states. In particular, we find selective disruption of interferon signaling in the most primitive HSCs of MDS patients. AZA partially rescued dysregulated gene expression and splicing in LR-MDS patient HSPCs. Shared splicing neoantigens observed in MDS represent novel immunotargets to overcome HMA resistance.

References

1 Zhang, X. et al. An immunophenotype-coupled transcriptomic atlas of human hematopoietic progenitors. Nat Immunol 25, 703-715, doi:10.1038/s41590-024-01782-4 (2024).

2 Li, G. et al. Splicing neoantigen discovery with SNAF reveals shared targets for cancer immunotherapy. Sci Transl Med 16, eade2886, doi:10.1126/scitranslmed.ade2886 (2024).