Quantitation of Leukemic Stem Cell-like Cells (LSC) Burden at Diagnosis Identifies Clinical Heterogeneity in the ELN22 Risk Stratification of Acute Myeloid Leukemia (AML)

Introduction: Despite advancements in the understanding of leukemia biology, AML remains a disease characterized by suboptimal outcomes. These suboptimal outcomes are driven by relapse within a significant proportion of patients. Relapse is thought to originate from chemo-resistant, dormant LSC populations that lack immunogenicity. Although evidence exists that the LSC burden at diagnosis may play an important prognostic role, current risk stratification schema for AML such as the European LeukemiaNet 2022 (ELN22) classification do not consider this. Here, we evaluate the LSC burden at diagnosis in each ELN22 risk category and demonstrate that this is an important predictor of outcome.

Methodology: A cohort of 150 patients of adult (≥ 18 years) AML were accrued at a single center over 8 years (2015 - 2023). All patients received uniform “3 + 7” induction chemotherapy. Patient characteristics, including age, sex, WBC count, and morphological remission status were evaluated. Assessment of measurable residual disease using multicolour flow cytometry (MFC-MRD) was performed at the post-induction (PI) time point. Diagnostic samples were sequenced using a 135-gene capture-based myeloid panel and risk-stratified as per ELN22 recommendations. Samples for baseline LSC analysis were processed using either a 10-color 2-tube assay (n=114) or a 16-color single-tube assay (n=36). Templates were standardized on MRD-negative BM samples of acute lymphoblastic leukemia patients. CD34+ CD38 dim to negative progenitors were analyzed using a combination of different from normal (DfN) and leukemia-associated immuno-phenotype (LAIP) approaches. We used a cut-off of 0.1% as previously described (Terwijn et al. Plos One 2014) to discriminate between LSC^high and LSC^low burden groups. Correlation of these groups with PI MFC-MRD and morphological remission status was performed using odds ratio. Overall survival (OS) and relapse-free survival (RFS) were used as the endpoints. Analysis of outcomes was performed using the log-rank test and Cox proportional hazard regression model.

Results: The median age of the cohort was 37 years (18-60 years) with male to female ratio of 1.7:1. The median follow-up was 34.8 months. The median overall survival (OS) was not reached whereas the median relapse-free survival (RFS) was 44.6 months. The median WBC count at diagnosis was 13.2✕10⁹/L (0.03-235.8). According to the ELN22 risk stratification, patients were classified as favorable (n=71, 47.3%), intermediate (n=57, 38%), and adverse (n=22, 14.6%) risk. The median LSC percentage of the 150 cases was 0.2% (0.001-55.9%). The majority of cases were classified as LSC^high (n=86, 57.3%). At the PI time point, 35 out of 150 patients were not in morphological remission, of which the majority (65%, n=23), had a LSC^high status. Additionally, 46.7% of the cases (n=70) were MFC-MRD positive, and 71.4% (n=50) of these cases were LSC^high. On univariate analysis, age, sex, and baseline WBC count had no impact on outcomes. The ELN22 adverse (HR=3.05, 95% CI 1.18-7.87, p=0.014) and intermediate-risk (HR=1.57, 95% CI 0.85-2.92) groups had significantly inferior OS, compared to favorable risk. A high LSC burden at diagnosis predicted for PI MFC-MRD positive status (p=0.0013). LSC^high AML had a significantly inferior OS (HR=2.10, 95% CI 1.15-3.82, p=0.01) and RFS (HR=2.55, 95% CI 1.31-4.96, p=0.004), compared to the LSC^low group. On analyzing individual ELN22 risk categories, LSC status did not impact OS and RFS in the favorable and adverse ELN22 risk groups. However, LSC status significantly impacted both RFS (HR=4.5, 95% CI 1.67-12.39, p=0.003) and OS (HR=2.9, 95% CI 1.23-7.20, p=0.017) in the ELN22 intermediate risk group. Furthermore, the median OS of the ELN22 intermediate group with LSC^high (15.89, 95% CI 12.73-30.79) was found to be similar to the ELN22 adverse risk group (16.61, 95% CI 10.33-19.70). On multivariate analysis, the ELN22 adverse risk group and LSC^high (0.1%) were found to be independent predictors of overall survival.

Conclusion: We demonstrate that the quantitation of LSC at diagnosis has an important role in predicting outcomes in AML. We further demonstrate that the ELN22 intermediate risk group with high baseline LSC load has outcomes comparable with the ELN22 adverse risk group.

Patkar:Illumina Inc: Research Funding.