Introduction Leukemic stem cells (LSCs) possess biological properties shared with normal hematopoietic stem cells. They confer resistance to chemotherapy and mediate relapse in acute myeloid leukaemia (AML) as well as in myelodysplastic syndromes (Ho et al., Blood 2016; Wang et al., Blood Advances 2020). A 17-gene stemness score ("LSC17") was predictive of treatment response and overall survival (OS) in AML (Ng et al., Nature 2016). Whilst chronic myelomonocytic leukaemia (CMML) is regarded as a myeloid stem cell neoplasm, the biological impact and prognostic implications of LSCs in CMML have not been explored. To this end, we evaluated the prognostic potential of these 17 genes in CMML patients.

Methods We performed mRNA sequencing on bone marrow (BM) samples from 44 CMML patients at presentation, each with extensive clinical and laboratory correlates and longitudinal outcome data. CMML diagnosis was based on The International Consensus Classification (Arber et al., Blood 2022). The expression of each LSC17 gene was evaluated and integrated with clinical and outcome metadata.

Results We identified three genes (MMRN1, ADGRG1, and DNMT3B) whose expression correlated significantly with overall survival (OS). An LSC-3 score was constructed based on the weighted sums derived from Cox regression analysis. Patients were divided into high and low LSC-3 score groups, with cut point determined by maximally selected rank statistics. Higher LSC-3 scores were associated with higher age (p=0.016), lower hemoglobin levels (p=0.011), and higher incidence of IDH1/2 mutations (p=0.009). High-score patients had significantly shorter leukaemia-free survival (LFS; median 7.4 months vs 30.5 months, p<0.001) and OS (10.5 months vs 33.9 months, p<0.001, Figure 1A). In subgroup analyses, LSC-3 score retained strong LFS and OS significance in both CMML-1 and CMML-2 stages as well as across dysplastic and proliferative CMML subgroups.

Mutations in ASXL1 have been consistently associated with poor prognosis in CMML and incorporated into several existing prognostic systems (Patnaik, Haematologica2022). In patients with unmutated ASXL1, high LSC-3 score robustly correlated with inferior LFS and OS (p=0.001 and p=0.001, respectively). In multivariable analysis, higher LSC-3 score remained an independent adverse risk factor for LFS and OS, irrespective of CPSS and CPSS-Molecular risk status. Time-dependent ROC curve analysis suggested potential for LSC-3 score to improve current risk stratification systems, increasing area under curve when incorporated (Figure 1B).

We validated the prognostic performance of the LSC-3 score in three independent CMML cohorts. Remarkably, LSC-3 score retained strong discriminatory performance even in a CMML monocyte RNA-seq dataset, suggesting potential application on peripheral blood samples.

GSEA revealed enriched oxidative stress, altered metabolism, and aberrant immune and T-cell functions in LSC-3 high BM samples. CIBERSORTx analysis (Newman et al., Nature Biotechnology 2019) also indicated significantly lower fractions of activated NK-cells (p=0.026) and borderline significant fewer CD8 T-cells (p=0.06) in the BM from high-score CMML patients. Finally, single-sample GSEA highlighted differences in drug sensitivity profiles, including resistance signatures for hypomethylating agents and venetoclax in LSC-3-high patients, which might partially explain the dismal prognosis in this population.

Discussion We describe a simple yet powerful three gene expression signature ("CMML-LSC3") as an independent and robust prognostic factor for LFS and OS in CMML patients. We validated this signature in three independent CMML cohorts with linked transcriptomic data, comprising both bone marrow mononuclear and sorted peripheral blood monocytes. We propose that LSC3 could be incorporated to complement existing scoring systems to improve prognostic performance and aid management decisions. Biologically LSC3 stratifies CMML patients by apparent differences in metabolism, immune microenvironment composition and drug resistance, raising the prospect of personalized therapeutic approaches mandating further study. The LSC3 score is being validated prospectively in a UK multicentre phase 2 trial (AMMO trial EudraCT 2021-004585-35).

Jerez:BMS: Honoraria; Novartis: Honoraria; Gilead: Research Funding. Somervaille:Abbvie: Consultancy; Bristol Myers Squibb: Consultancy; Novartis: Consultancy; Imago Biosciences: Research Funding; Oryzon Genomics: Consultancy; CellCentric Ltd: Research Funding. Chou:aether AI: Other: The AI product in this abstract... investigators of this study w/commercial profit; National Taiwan University Hospital: Other: The AI product in this abstract... investigators of this study w/commercial profit. Wiseman:Novartis: Honoraria, Membership on an entity's Board of Directors or advisory committees; Jazz Pharmaceuticals: Honoraria; Astex Pharmaceuticals: Research Funding; Gilead Sciences: Membership on an entity's Board of Directors or advisory committees; Blueprint Oncology: Membership on an entity's Board of Directors or advisory committees.

Author notes


Asterisk with author names denotes non-ASH members.

Sign in via your Institution