对AML PDX模型的整合转录组分析和流式细胞术分析鉴定出LSCs与原始细胞共有的嘌呤-嘧啶代谢依赖性

Acute myeloid leukemia (AML) comprises a heterogeneous population in which leukemic stem cells (LSCs) can drive initiation, persistence, and relapse. While current standard of care effectively reduces disease burden, 30-80% relapse rates emphasize the need of investigating LSC as a potential driver of relapse. To better understand biology of LSC, we performed RNA-seq and flow cytometry analysis in AML patient‑derived xenograft (PDX) models with defined LSC, progenitor, and blast compartments. Ex vivo culture conditions were optimized for AML PDXs to preserve subpopulation heterogeneity in vitro by supplementing cytokines. For RNA-seq, CD34/CD38 markers-based sorting isolated LSCs, progenitors, and blasts; differential expression analyses identified LSC‑specific genes and shared vulnerabilities that could enable dual targeting of LSCs and blasts. For immunophenotyping, we developed a 20‑color flow cytometry panel incorporating LSC, differentiation, and classic AML surface markers to quantify expression patterns across compartments. RNA-seq of sorted LSCs and blasts demonstrated differential expression between LSCs and blasts with a clear upregulation in the LSC17 score, consistent with previous results and confirming our data purity/quality. Flow cytometry panel also displayed differential surface marker expression across LSCs and blasts. Focusing on LSC-enriched biology first, we identified around 160 genes consistently upregulated in LSCs across 2 independent AML PDX models. These LSC-elevated targets showed limited dependency in DepMap (which have AML cell lines that represent blast), suggesting these genes may exert functional effects only in LSC, and not in blasts. To identify targets showing a dependency in all subsets, we implemented a blast-centric filter. We first selected AML dependency genes from DepMap and then intersected this set with genes highly expressed in both LSCs and blasts. This approach yielded 16 candidate genes with dual relevance. Functional annotation indicated that half of these genes map to purine/pyrimidine biosynthetic pathways, pointing to a metabolic vulnerability and supporting a potential dual-targeting strategy to impact both the LSC and blast population. Integrated transcriptomic profiling and flow cytometry analysis exhibit LSC-restricted programs while uncovering shared metabolic dependencies across LSCs and blasts. By combining expression with AML dependency data, we prioritized 16 dual-compartment candidates, where about half align to purine/pyrimidine biosynthesis nominating nucleotide metabolism as a potential axis for dual targeting. Together elevated LSC17 scores that confirm dataset quality, these findings provide a roadmap for target selection, patient stratification, and rational combinations aimed at durable remission and reduced relapse in AML.