Acute myeloid leukemia (AML) is sustained by a subpopulation of rare leukemia-initiating cells (LIC) detected in the xenograft assay by their capacity to self-renew and to generate non-LICs in vivo. The xenotransplantation model captures functional properties of LICs that have clinical prognostic value. However, the long duration of this in vivo assay has hampered its use as a prognostic tool. Here, we show, using an ex vivo coculture system, that intermediate and poor risk AML patient samples at diagnosis have a 5 to 7 times higher frequency of leukemic long-term culture-initiating cells (L-LTC-IC) compared with the good risk group. We defined a fluorescence dilution factor (FDF) parameter that monitors sample proliferation over 1 week and established a strong correlation of this parameter with the L-LTC-IC frequency. A higher FDF was found for poor prognostic AMLs or for samples capable of engrafting NSG mice compared with good risk AMLs or nonengrafters. Importantly, FDF could classify normal karyotype intermediate risk patients into two groups with a significant difference in their overall survival, thus making this nongenetic and non-in vivo approach a new clinically relevant tool for better diagnosis of AML patients. Cancer Res; 76(8); 2082–6. ©2016 AACR.

Leukemia-initiating cells (LIC) are functionally defined as SCID leukemia–initiating cells (SL-IC; ref.1) in the xenograft assay by their capacity to initiate, propagate, and maintain bulk leukemia in vivo (2). Functional studies of SL-ICs showed a correlation between the xenograft capacity of a sample as well as “stem cell gene signature”, with poorer overall survival (OS) of the respective patient (3, 4).

In addition, an acute myeloid leukemia (AML) mathematical modeling of LIC proliferation was also separately shown to correlate with the clinical outcome of the patients (5). Thus, LIC quantification and their monitoring could have strong clinical applications, especially for intermediate-risk normal karyotype AMLs that account for approximately 60% of all AML patients. However, the recently described heterogeneous SL-IC phenotypes (4, 6–8) combined with the long duration of the in vivo assay have prevented the use of the xenograft assay as a prognostic tool. We have recently optimized a niche-like coculture system capable of maintaining SL-IC ex vivo and demonstrated that the frequency of leukemic long-term culture-initiating cells (L-LTC-IC) is a reliable functional readout for monitoring the activity of LICs (9). Here, we combine this assay with a cell proliferation analysis to demonstrate that the expansion rate of L-LTC-ICs in this culture system strongly correlates with patient clinical outcome.

Cells

AML cells were obtained at St Bartholomew's Hospital (London, United Kingdom) and from the Institutional Tumor Bank at Institut Paoli-Calmettes (Comprehensive Cancer Centre, Marseille, France). For both sources, ethical approvals have been granted (via the East London Ethical Community or under authorization #AC-2013-1905 granted by the French Ministry of Research, respectively). Details of patient samples are listed in Supplementary Table S1. Coculture experiments were performed as described previously (9) on confluent MS-5 monolayers. The stromal cell line MS-5 was purchased from German Collection of Microorganisms and Cell Cultures (http://www.dsmz.de) in 2012 and were maintained in IMDM 10% FCS + 2 mmol/L l-glutamine and used between passage 3 to 5.

Fluorescence dilution factor

AML cells were stained with 0.8 μmol/L carboxyfluorescein diacetate succinimidyl ester (CFSE; Invitrogen). Cells were washed and incubated on preestablished confluent MS-5. CFSE median fluorescence intensity (MFI) was measured by FACS at 18 hours and day 7 on viable (Annexin V and DAPI negative) human hematopoietic cells (CD45+ and Sca-1 negative). Fluorescence dilution factor (FDF) was defined as the ratio of the 18-hour CFSE MFI divided by the 1-week CFSE MFI (See Supplementary Material for more details).

Statistical analysis

Data were analyzed for statistical significance using the Mann–Whitney unpaired two-tailed test or the one-way ANOVA test. Linear or nonlinear regression trend lines were performed with GraphPad Prism software. A nonparametric Spearman test was applied for correlation. Spearman rank correlation coefficient (ρ) is shown. Observed differences were regarded as statistically significant if the calculated two-sided P value was below 0.05.

For additional details, see Supplementary Material and Methods.

We analyzed 92 de novo AML patients classified as favorable (n = 22), intermediate (n = 54), and poor (n = 16) prognostic groups according to the British MRC and French BGMT classifications (Supplementary Table S1). We performed ex vivo limiting dilution analyses to determine the initial frequencies of L-LTC-IC 1° frequency (Fig. 1A). We observed a high variability ranging from 1 L-LTC-IC in 10 to 105 bulk AML-plated cells and noticed that the intermediate and poor risk AMLs had a 5 to 7 times higher frequency as compared with the favorable group (Fig. 1B, left). Cell counts at 5 weeks were found to be different depending on AML risk groups (Fig. 1B, right). Importantly, for all samples, we also correlated the L-LTC-IC 1° frequency (Fig. 1C) and the 5-week fold expansion (Fig. 1D) with the patients' OS. By plotting the L-LTC-IC 1° frequencies against the 5-week fold expansion, we further confirmed that the very modest ex vivo proliferation of AMLs depends on the L-LTC-IC compartment size of the sample (Fig. 2A). This correlation was not seen after a 1-week culture period (Fig. 2B). This suggests either that L-LTC-ICs did not sustain the leukemic expansion during the first week or more likely that cell death exceeded the L-LTC-IC–driven leukemic expansion. To address this question, we quantified secondary L-LTC-ICs (LTC-IC 2°) after replating cells that have been cultured for one week (Fig. 1A). We observed that the median proportion of L-LTC-ICs increased 7.8 times on average (Fig. 2C, left; n = 42, P < 0.001). In parallel, we quantified the loss of cellularity during the first week to be 60%, suggesting that the 7.8 times increase seen in L-LTC-ICs proportion was due to the ongoing proliferation of some LICs. This was confirmed by calculating the L-LTC-IC absolute count, which increases by a factor 4.25 (n = 16, P < 0.05; Fig. 2C right). Consistently, the intermediate and poor risk samples maintained their higher proportion of L-LTC-ICs as compared with the favorable risk group after replating (Fig. 2D). Thus, L-LTC-IC self-renewal capacity is an intrinsic biologic feature of samples that can be monitored over a 1-week culture period.

Figure 1.

AML long-term culture potential correlates with patient prognostic risk and clinical outcome. A, flow chart illustrating the experimental design for B–D. B, frequency (Freq) of L-LTC-IC in primary plating (L-LTC-IC 1°) for good, intermediate/normal karyotype, and poor risk AML patients (left) and the fold expansion of AML population after 5 weeks of coculture with MS-5 for good, intermediate/normal karyotype, and poor risk AML patients (right). C and D, L-LTC-IC 1° frequency (C) or 5-week AML fold expansion (D) as compared with patient's OS (n = 45). Black line shows experimentally derived nonlinear regression trend line. Mann–Whitney unpaired two-tailed test or the one-way ANOVA test was applied. *, P < 0.05; **, P < 0.01; NS, not significant, P > 0.05. LDA, limiting dilution assay.

Figure 1.

AML long-term culture potential correlates with patient prognostic risk and clinical outcome. A, flow chart illustrating the experimental design for B–D. B, frequency (Freq) of L-LTC-IC in primary plating (L-LTC-IC 1°) for good, intermediate/normal karyotype, and poor risk AML patients (left) and the fold expansion of AML population after 5 weeks of coculture with MS-5 for good, intermediate/normal karyotype, and poor risk AML patients (right). C and D, L-LTC-IC 1° frequency (C) or 5-week AML fold expansion (D) as compared with patient's OS (n = 45). Black line shows experimentally derived nonlinear regression trend line. Mann–Whitney unpaired two-tailed test or the one-way ANOVA test was applied. *, P < 0.05; **, P < 0.01; NS, not significant, P > 0.05. LDA, limiting dilution assay.

Close modal
Figure 2.

L-LTC-IC self-renew during the first week of culture with amplitude related to patient prognosis. For schematic, see the flow chart in Fig 1A. A and B, five-week or 1-week AML fold expansion plotting against L-LTC-IC 1° frequency (n = 43 and 22, respectively). A, thin black line shows experimental-derived nonlinear regression trend line with 95% confidence band (dashed lines). For A and B, a nonparametric Spearman correlation test was applied. C, frequency (left; n = 73) and absolute count (right; n = 16) of L-LTC-IC in primary and secondary plating. D, frequency of L-LTC-IC in secondary plating (L-LTC-IC 2°) for good, intermediate/normal karyotype, and poor risk AML samples. Mann–Whitney unpaired or paired two-tailed test (C) or the one-way ANOVA test (D) was applied. *, P < 0.05; **, P < 0.01; ***, P < 0.001; NS, not significant, P > 0.05. LDA, limiting dilution assay.

Figure 2.

L-LTC-IC self-renew during the first week of culture with amplitude related to patient prognosis. For schematic, see the flow chart in Fig 1A. A and B, five-week or 1-week AML fold expansion plotting against L-LTC-IC 1° frequency (n = 43 and 22, respectively). A, thin black line shows experimental-derived nonlinear regression trend line with 95% confidence band (dashed lines). For A and B, a nonparametric Spearman correlation test was applied. C, frequency (left; n = 73) and absolute count (right; n = 16) of L-LTC-IC in primary and secondary plating. D, frequency of L-LTC-IC in secondary plating (L-LTC-IC 2°) for good, intermediate/normal karyotype, and poor risk AML samples. Mann–Whitney unpaired or paired two-tailed test (C) or the one-way ANOVA test (D) was applied. *, P < 0.05; **, P < 0.01; ***, P < 0.001; NS, not significant, P > 0.05. LDA, limiting dilution assay.

Close modal

We next wondered whether a shorter and simpler test, using CFSE staining to track cell division, could be implemented to stratify sample risk groups. However, a high resolution of division peaks could not be achieved for the majority of samples due to intrasample morphologic heterogeneity, as the incorporation of CFSE depends on cell size (for more details, see Supplementary Material and Methods and Supplementary Fig. S1A–S1C). Moreover, we could not use cell sorting to reduce biologic heterogeneity without the possibility of biasing the cell population studied (10). As AML samples usually have a poor viability at thawing and at later points (see details of influence of cell viability in Supplementary Figs. S2A–S4C), we measured the median dye dilution of the nonapoptotic leukemic cell population (non-DAPI, non-Annexin–positive fraction, Supplementary Fig. S2B and S2C). We then defined the FDF parameter as the ratio of MFI at the start of the analysis by the MFI measured after 1 week of coculture (Fig. 3A and Supplementary Fig. S3). Heterogeneous FDF values ranging from 1 to 10.2 with a mean of 3.0 were determined for 80 AML samples. We observed a strong correlation between FDF and the L-LTC-IC frequency (Fig. 3B) or the proliferation index (Supplementary Fig. S4A), strongly supporting the notion that the FDF reports leukemic stem/progenitors dynamics ex vivo. FDF did not correlate with bulk leukemic expansion or bulk sample viability (Supplementary Fig. S4B and S4C). Thus, the FDF parameter is informative of subpopulation dynamics even when cell death over exceeds proliferation. Of note, the FDF parameter cannot distinguish between a high proportion of slowly dividing cells versus a small fraction of highly dividing ones.

Figure 3.

One-week FDF correlates with L-LTC-IC and predicts E versus NE in NSG mice. A, flow chart illustrates the experimental procedure to define the FDF values (top; see Supplementary Fig. S2 for FACS gating and cytometer calibration strategies) and representation of 18-hour and 1-week overlay CFSE fluorescence profiles of one patient with high FDF (bottom left) and one patient with low FDF value (bottom right). B, FDF plot against the L-LTC-IC frequency (n = 36). Thin black line shows experimental derived nonlinear regression trend line. Nonparametric Spearman correlation test was applied. C, FDF value between NSG mice E and NE AML samples (E, n = 18; NE, n = 14). *, P < 0.05; ***, P < 0.001. NS, not significant. D, fold expansion at 1 week for E and NE samples.

Figure 3.

One-week FDF correlates with L-LTC-IC and predicts E versus NE in NSG mice. A, flow chart illustrates the experimental procedure to define the FDF values (top; see Supplementary Fig. S2 for FACS gating and cytometer calibration strategies) and representation of 18-hour and 1-week overlay CFSE fluorescence profiles of one patient with high FDF (bottom left) and one patient with low FDF value (bottom right). B, FDF plot against the L-LTC-IC frequency (n = 36). Thin black line shows experimental derived nonlinear regression trend line. Nonparametric Spearman correlation test was applied. C, FDF value between NSG mice E and NE AML samples (E, n = 18; NE, n = 14). *, P < 0.05; ***, P < 0.001. NS, not significant. D, fold expansion at 1 week for E and NE samples.

Close modal

Most good risk AML samples are devoid of xenograft potential (4). Among our cohort, 32 samples were tested for their ability to propagate AML in a xenograft assay in NSG mice. We compared the FDF values of NSG mice engrafter (E) and nonengrafter (NE) AML samples. A higher FDF was found for 18 E compared with 14 NE samples (Fig. 3C), which further suggests that the FDF index reports on the actual LICs dynamics ex vivo. On the other hand, no significant differences were seen between the E/NE groups simply using fold cell expansion as the parameter (Fig. 3D). Fourteen of the 19 good risk samples had a FDF below the mean value of 3.0 of this cohort while 11 of the 15 poor risk AMLs tested were above the mean (Fig. 4A). Intermediate risk normal karyotype AML samples were stratified across the whole range. Poor prognostic AMLs had a significantly higher FDF values compared with good risk and intermediate risk group AMLs, whereas no statistical differences were seen between good risk and intermediate risk groups (Fig. 4B). Again, using cell expansion as parameter, we were unable to detect any differences between the 3 groups (Fig. 4C). Importantly, patients' OS was significantly reduced in AML with high FDF (>3.0) when compared with AML cases with low FDF (<3.0, Fig. 4D, left). As the intermediate risk group contains patients with variable outcomes, we evaluated the usefulness of the FDF index to correlate with their outcome. Using the same cut-off value, we could divide the patient cohort and show for 17 AMLs with high FDF a statistically significantly lower survival compared with 27 AMLs with low FDF (Fig. 4D, second panel from the left). Recently, the combined mutational status of FLT3 and NPM1 has been found to stratify intermediate risk/normal karyotype group in low molecular risk (NPM1mut FLT3wt) intermediate 1 and high molecular risk (FLT3-ITD or NPM1wt FLT3wt) intermediate 2 groups (11). High FDF negatively correlated with a lower OS in these two subgroups (Fig. 4D, first and second panel from the right). FDF measurements could also refine good and poor risk groups, although the survival curve between low and high FDF was not significant based on the small outlier number of patients in these two groups (Supplementary Fig. S5).

Figure 4.

FDF predicts clinical outcome. A, FDF for 80 AML samples. Poor risk (red bars), n = 15; intermediate/normal karyotype (orange bars), n = 46; good risk (green bars), n = 19. Black line shows the mean FDF value (=3.04). Dashed line represents FDF = 1, no dilution from the input fluorescence. B, FDF value comparison for different risk group AML. C, 1-week fold expansion at 1 week for different risk group. D, patients with low FDF have better OS than high FDF patients. Kaplan–Meier 5-year survival curves based on the mean FDF cut-off value determined in A. Patients who underwent an allograft were excluded from the analysis. Left, all risk group; second panel, intermediate risk normal karyotype (NK) group; third panel, intermediate 1 group, NPM1mut/FLT3wt cytogenetically normal AML; fourth panel, intermediate 2, NPM1wt or FLT3ITD cytogenetically normal AML. Mantel–Cox log-rank test was applied for D. One-way ANOVA test was applied for B and C. *, P < 0.05; NS (not significant), P > 0.05.

Figure 4.

FDF predicts clinical outcome. A, FDF for 80 AML samples. Poor risk (red bars), n = 15; intermediate/normal karyotype (orange bars), n = 46; good risk (green bars), n = 19. Black line shows the mean FDF value (=3.04). Dashed line represents FDF = 1, no dilution from the input fluorescence. B, FDF value comparison for different risk group AML. C, 1-week fold expansion at 1 week for different risk group. D, patients with low FDF have better OS than high FDF patients. Kaplan–Meier 5-year survival curves based on the mean FDF cut-off value determined in A. Patients who underwent an allograft were excluded from the analysis. Left, all risk group; second panel, intermediate risk normal karyotype (NK) group; third panel, intermediate 1 group, NPM1mut/FLT3wt cytogenetically normal AML; fourth panel, intermediate 2, NPM1wt or FLT3ITD cytogenetically normal AML. Mantel–Cox log-rank test was applied for D. One-way ANOVA test was applied for B and C. *, P < 0.05; NS (not significant), P > 0.05.

Close modal

The advance of next-generation sequencing for screening mutations in AML patients has improved patients' stratification; nevertheless, this screening is not yet a bedside standardized procedure. On the other hand, the quantification of LICs as well as the presence of a “stem cell signature” has been shown to provide information on the clinical outcome of patients but is hard to use in a routine setting. Here, our data demonstrate that the ex vivo frequency of L-LTC-IC and its expansion dynamics reflect the intrinsic biology of the LICs. We further show that monitoring AML culture–initiating cells expansion after 1 week could help predict the prognosis of AML patients without the need of in vivo experiments. Here, our data demonstrate that the ex vivo frequency of L-LTC-IC and its expansion dynamics reflects the intrinsic biology of LICs. We further show that monitoring AML culture–initiating cell expansion after 1 week could help predict the prognosis of AML patients without the need of in vivo experiments.

No potential conflicts of interest were disclosed.

Conception and design: E. Griessinger, F. Anjos-Afonso, D.C. Taussig, J.G. Gribben, D. Bonnet

Development of methodology: E. Griessinger, F. Anjos-Afonso

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Griessinger, F. Anjos-Afonso, J. Vargaftig, F. Lassailly, T. Prebet, V. Imbert, M. Nebout, N. Vey, C. Chabannon, F. Bollet-Quivogne

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Griessinger, F. Anjos-Afonso, F. Lassailly, T. Prebet, V. Imbert, A. Filby, D. Bonnet

Writing, review, and/or revision of the manuscript: E. Griessinger, F. Anjos-Afonso, D.C. Taussig, T. Prebet, N. Vey, C. Chabannon, F. Bollet-Quivogne, J.G. Gribben, J.-F. Peyron, D. Bonnet

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Anjos-Afonso, D. Bonnet

Study supervision: E. Griessinger, J.G. Gribben, D. Bonnet

The authors thank the patients who granted permission to use their samples for research, Finlay McDougall for providing diagnostic information, and all personnel at the Institut Paoli-Calmettes Tumour Bank for the access of anonymized samples and clinical data. The authors also thank Stuart Horswell for statistical analysis and Dr. Katie Foster for proofreading the manuscript.

E. Griessinger was supported by Cancer Research UK internal fellowship and by a grant from the Fondation de France. T. Prebet, N. Vey, and C. Chabannon were supported by grant INCa-DGOS-Inserm 6038 to the SIRIC PACA-Ouest. D. Bonnet was funded by Cancer Research UK and by European grant (contract no.: 037632). J.F Peyron was supported by INSERM and by a grant from the Cancéropôle PACA.

1.
Bonnet
D
,
Dick
JE
. 
Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell.
Nat Med
1997
;
3
:
730
7
.
2.
Dick
JE
. 
Stem cell concepts renew cancer research.
Blood
2008
;
112
:
4793
807
.
3.
Pearce
DJ
,
Taussig
D
,
Zibara
K
,
Smith
LL
,
Ridler
CM
,
Preudhomme
C
, et al
AML engraftment in the NOD/SCID assay reflects the outcome of AML: implications for our understanding of the heterogeneity of AML.
Blood
2006
;
107
:
1166
73
.
4.
Eppert
K
,
Takenaka
K
,
Lechman
ER
,
Waldron
L
,
Nilsson
B
,
van Galen
P
, et al
Stem cell gene expression programs influence clinical outcome in human leukemia.
Nat Med
2011
;
17
:
1086
93
.
5.
Stiehl
T
,
Baran
N
,
Ho
AD
,
Marciniak-Czochra
A
. 
Cell division patterns in acute myeloid leukemia stem-like cells determine clinical course: a model to predict patient survival.
Cancer Res
2015
;
75
:
940
9
.
6.
Taussig
DC
,
Miraki-Moud
F
,
Anjos-Afonso
F
,
Pearce
DJ
,
Allen
K
,
Ridler
C
, et al
Anti-CD38 antibody-mediated clearance of human repopulating cells masks the heterogeneity of leukemia-initiating cells.
Blood
2008
;
112
:
568
75
.
7.
Taussig
DC
,
Vargaftig
J
,
Miraki-Moud
F
,
Griessinger
E
,
Sharrock
K
,
Luke
T
, et al
Leukemia-initiating cells from some acute myeloid leukemia patients with mutated nucleophosmin reside in the CD34(-) fraction.
Blood
2010
;
115
:
1976
84
.
8.
Sarry
JE
,
Murphy
K
,
Perry
R
,
Sanchez
PV
,
Secreto
A
,
Keefer
C
, et al
Human acute myelogenous leukemia stem cells are rare and heterogeneous when assayed in NOD/SCID/IL2Rgammac-deficient mice.
J Clin Invest
2011
;
121
:
384
95
.
9.
Griessinger
E
,
Anjos-Afonso
F
,
Pizzitola
I
,
Rouault-Pierre
K
,
Vargaftig
J
,
Taussig
D
, et al
A niche-like culture system allowing the maintenance of primary human acute myeloid leukemia-initiating cells: a new tool to decipher their chemoresistance and self-renewal mechanisms.
Stem Cells Transl Med
2014
;
3
:
520
9
.
10.
Begum
J
,
Day
W
,
Henderson
C
,
Purewal
S
,
Cerveira
J
,
Summers
H
, et al
A method for evaluating the use of fluorescent dyes to track proliferation in cell lines by dye dilution.
Cytometry A
2013
;
83
:
1085
95
.
11.
Dohner
H
,
Estey
EH
,
Amadori
S
,
Appelbaum
FR
,
Büchner
T
,
Burnett
AK
, et al
Diagnosis and management of acute myeloid leukemia in adults: recommendations from an international expert panel, on behalf of the European LeukemiaNet.
Blood
2010
;
115
:
453
74
.