High Leukemic Stem Cell Frequency in Remission Bone Marrow Predicts Poor Outcome in Acute Myeloid Leukemia

Relapses of acute myeloid leukemia (AML) originate from the outgrowth of therapy surviving leukemic blasts known as minimal residual disease (MRD). We and others have shown that immunophenotypical detection of MRD in remission bone marrow (BM) is a strong prognostic factor for survival. However, some patients relapse despite low MRD levels, indicating the presence of remaining, but undetectable leukemic stem cells (LSCs). These LSCs are proposed to give rise to AML and because we previously found a correlation between stem cells at diagnosis and MRD (van Rhenen et al., CCR 2005), they might be responsible for relapse. Therefore, we investigated whether LSCs could be detected in remission BM and whether monitoring remaining LSCs might predict clinical outcome more accurately as compared to the immunophenotypically defined “total blast” MRD. We defined LSCs by FACS analysis by the use of aberrant lineage markers, as well as the novel LSC marker CLL-1 (van Rhenen et al., Leukemia 2007 and Blood 2007) in 22 patients with CD34+ leukemia. In patients with more than 20% leukemia associated phenotype (LAP) positive CD34+ CD38− stem cells, 37 follow-up remission BM samples were collected (17 after first cycle, 13 after second cycle and 7 after third cycle of chemotherapy) and analysed for LAP expression. A stem cell compartment was defined as a minimum of 5 clustered CD34+ CD38− events when a minimum of 500,000 white blood cells (WBCs) were measured. In cases of less than 5 events, LSC frequency was set at <1/500,000 (< 2 × 10⁻⁴) LSCs/total WBC. In 10 follow-up samples, no CD34+CD38− compartment could be detected. In the other samples, the markers CD7 (n=9), CD11b (n=2), CD19 (n=3), CD56 (n=1) and CLL-1 (n=12) were used to detect LSCs. In addition, in 4 of these cases, CD45 expression, cell size (forward light scatter, FSC) and cell granularity (sideward light scatter, SSC) characteristics were used as a secondary gating strategy to exclude malignant cells in the aberrant marker negative fraction (M.Terwijn, accompanying abstract for this meeting). After the first cycle of chemotherapy (n=17), high LSC frequency (> 1 × 10⁻³) clearly predicted overall survival (OS), relapse free survival (RFS) and disease free survival (DFS). Regarding OS, LSC frequency above this cut-off lead to a median survival of 8 months (n=7) vs. >29 months in the group above cut-off (n=10, p=0.028). Median RFS in the high LSC frequency group was 8 months vs. >27 months (p=0.019) and for DFS, these numbers were 6 months vs. >27 months, respectively (p=0.025). After the second and third cycle of chemotherapy a cut-off of 2 × 10⁻⁴ was used. Again, a high LSC number predicted OS, RFS and DFS (p=0.035, 0.029 and 0.029, respectively) with a median RFS of 6 months (n=6) versus >30 months (n=7) in patients with a low LSC frequency. Although only limited number of patients was followed after the third cycle, trends were similar: 3/3 patients with a high LSC frequency relapsed within 6 months after CR, while only 1/4 patients with a low LSC frequency relapsed within 6 months. The remaining 3 patients had a RFS of 11, 21 and 31 months. In contrast to LSC frequency, “total blast” MRD frequency did not predict prognosis after the first, second or third cycle of chemotherapy, which likely is due to the small sample size. LSC frequency is superior in predicting prognosis of AML patients in CR as compared to MRD total blast frequency. The prognostic value of remaining LSCs supports the hypothesis that the leukaemia initiating capacity in patients with AML originates from this population. The identification of LSCs in remission BM and the possibility to distinguish these from HSCs will allow monitoring differential effects of present and future therapies.

This work was supported by Netherlands Cancer Foundation KWF.