Flow cytometry-based assessment of mitoxantrone efflux from leukemic blasts varies with response to induction chemotherapy in acute myeloid leukemia


  • How to cite this article: Kim HP, Bernard L, Berkowitz J, Nitta J, Hogge DE. Flow cytometry-based assessment of mitoxantrone efflux from leukemic blasts varies with response to induction chemotherapy in acute myeloid leukemia. Cytometry Part B 2012; 82B: 283–294.



Accurate prediction of chemotherapy drug resistance would aid treatment decisions in acute myeloid leukemia (AML). The aim of this study was to determine if mitoxantrone efflux from AML blasts would correlate with response to induction chemotherapy.


Flow cytometry was used to measure the median fluorescence intensity (MFI) for AML blasts incubated with mitoxantrone [an ATP-binding cassette (ABC) transporter substrate] with or without coincubation with cyclosporine A (a broad-spectrum inhibitor of ABC transporters) and a ratio (MFIR) between the inhibited and uninhibited MFI was calculated.


Among 174 AML patient blast samples, the mean MFIR for complete remission (CR) patients was lower than that obtained for induction failure (IF) patients (mean MFIR ± SD 1.62 ± 0.53 for CR after one cycle of chemotherapy vs. 2.22 ± 1.29 for CR after two cycles and 2.59 ± 0.98 for IF, P < 0.001). Logistic regression analysis determined 2.45 as the MFIR threshold above which 29% of patients achieved CR vs. a CR rate of 84% when the MFIR was ≤ 2.45 (P < 0.0001). In AML patients with normal karyotype (n = 80), CR was obtained for 33% of patients with an MFIR > 2.45 vs. 89% of those with MFIR ≤ 2.45 (P < 0.0001). In patients > age 60 (n = 77), 30% vs. 87% of those with MFIR > vs. ≤ 2.45 achieved CR (P < 0.0001).


This assay of ABC transporter function can potentially predict response to induction chemotherapy in AML. © 2012 International Clinical Cytometry Society


Although most patients with acute myeloid leukemia (AML) achieve complete remission (CR) with conventional induction chemotherapy, at least 20% of younger patients and as many as 50% of those above 60 years have leukemia that is chemotherapy refractory at diagnosis (1, 2). Conventional AML induction chemotherapy typically consists of cytarabine in combination with an anthracycline in doses sufficient to induce temporary marrow aplasia (3). This therapy causes considerable morbidity and occasional mortality particularly in elderly patients and those with comorbid illness. If CR is not obtained with initial induction, survival is typically short especially in those not fit enough to undergo even more aggressive salvage treatment. A number of factors have been shown to have prognostic impact in AML. The most powerful of these are cytogenetic and/or molecular abnormalities detected in the diagnostic bone marrow (BM) sample (4, 5). However, 40%–50% of AML patients display a normal karyotype (NK) at diagnosis. Although NK AML is considered to have an intermediate prognosis overall, a wide variability in response to induction therapy and survival is seen in this group (6, 7). A number of molecular abnormalities with prognostic significance are now routinely assessed in the NK group (8). These include mutations in the fms-like tyrosine kinase 3 (FLT3), nucleophosmin 1 (NPM1), and CCAAT/enhancer binding protein α (CEBPA) genes (9–11). However, some NK patients have none of these abnormalities. Detection of cytogenetic and/or molecular abnormalities is extremely useful for planning of postremission therapy. However, in many centers, the results of this relatively labor intensive testing is not available for a number of days after the diagnosis of AML has been made and initial induction therapy should be initiated. The ability to predict the presence of chemotherapy refractory AML at diagnosis would be clinically useful for many patients (12). For example, the overall results of induction chemotherapy for AML patients above age 70 are very poor (13). If the futility of induction chemotherapy could be predicted, it could be avoided along with the associated toxicity. On the other hand, if achievement of CR was predicted to be highly likely elderly patients or those with significant comorbidities could be somewhat reassured that the risk associated with therapy should be justified by the benefit obtained (14–16).

As discussed above, resistance to conventional chemotherapy drugs, whether primary or acquired, is a major obstacle to successful treatment and a main cause of death in AML. There are 49 human genes in the ATP-binding cassette (ABC) transporter superfamily encoding transmembrane proteins, which mediate drug efflux via ATP-dependent pumps (17–21). Among the ABC transporters, expression of multidrug resistance protein 1 (MDR1 or P-glycoprotein) and breast cancer resistance protein 1 (BCRP1) have been associated with chemotherapy resistance in AML in some studies (22–28) although not in all (29–32). One possible reason for the conflicting prognostic relevance suggested for ABC transporter expression in AML is the variety of methods used to assess that expression in previous studies (33). The results of molecular or immunological testing would not necessarily be comparable with each other or to assays measuring transporter function. Testing which assesses ABC transporter function by measuring drug efflux allows the simultaneous assessment of the activity of more than one of these proteins, which may provide an advantage in predicting drug resistance over more targeted assessment of individual genes or proteins (6, 34).

Previous work from our center used quantitative reverse transcriptase polymerase chain reaction (QRT-PCR) to study MDR1 and BCRP1 expression in AML blasts and subpopulations of AML cells highly enriched for candidate leukemic stem cells (LSC) (35). The latter were studied as the persistence of LSC is thought to be a mechanism of chemotherapy refractoriness in AML (36). In this study, expression of both MDR1 and BCRP1 was higher in AML cells from chemotherapy refractory patients than in cells from those who achieved CR when the CD34+CD38 fraction enriched for LSC was studied (37). Although this data strongly suggested the prognostic relevance of quantifying ABC transporter expression in AML, the technique used required both cell sorting and QRT-PCR making it unlikely that a clinical laboratory would adopt this for routine testing of AML patients. Thus, the aim of this study was to develop a flow cytometry-based assay that would predict response to induction chemotherapy in newly diagnosed AML patients with a rapid turnaround that would allow the results to be available before initial treatment decisions are necessary. Flow cytometry was chosen because the technology is widely available in clinical hematology laboratories and routinely used for immunophenotyping of AML blasts at diagnosis. In the assay described, the median fluorescence intensity (MFI) for AML blasts incubated with mitoxantrone (an ABC transporter substrate) is measured with or without coincubation with cyclosporine A (a nonspecific inhibitor of ABC transporter function). A ratio between the inhibited and uninhibited MFI is then calculated (MFIR). This assay was applied to both the total blast cell population and the CD34+CD38 LSC-enriched fraction from newly diagnosed patients for whom the outcome of induction therapy was known. In this preliminary evaluation, the MFIR in AML blasts was higher in induction failure (IF) than in CR patients and it seemed possible that an MFIR value could be determined, which would predict the outcome from initial induction chemotherapy in AML patients including those with NK and above 60 years.


AML Patients and Samples

Peripheral blood (PB) or BM leukemia blast cells were obtained at diagnosis from newly diagnosed AML patients. All samples were obtained after informed consent and with the approval of the Clinical Research Ethics Board of the University of British Columbia. The diagnosis of AML was established on presentation BM biopsy and aspirate using World Health Organization criteria (38). Cytogenetic risk was assigned using MRC (UK) criteria (39). Twenty-eight initial samples (14 CR and 14 IF) were used in initial studies to evaluate expression of MDR1 and BCRP1 by real time RT-PCR and to compare this with the results of MFIR testing (Figs. 1, 2, 3, 4, 1–4). Three more AML samples arriving directly from the clinic were used to compare MFIR results on fresh vs. cryopreserved cells (Table 2). All other analysis was performed on a large cohort of samples from 174 patients for whom clinical characteristics are shown in Table 1.

Figure 1.

Representative FACS plots for MFIR analysis. Total PB or BM MNCs are stained with CD34/CD38 and CD3/CD19 antibodies and mitoxantrone as described in the Materials and methods section. Dead cells (PI positive) are excluded (A). Total (PI negative) MNCs or cells in the CD34+CD38- or CD3-CD19-gates are used for MFIR analysis (B).

Figure 2.

mRNA expression levels for MDR1 (A) and BCRP1(B) in AML patient samples. Expression levels in total PB MNCs (blasts), CD34+CD38+, and CD34+CD38 cells from 14 CR (•) and 14 NR (▪) patients were determined using QRT-PCR for MDR1 (A) and BCRP1 (B). Expression levels are set relative to GAPDH (set at 106). Horizontal lines represent the mean value for each group. P-values comparing CR and NR samples were determined using the Student t test.

Figure 3.

Correlation between QRT-PCR for ABC transporter expression and the MFIR. mRNA expression values as determined by QRT-PCR (normalized to GAPDH set at 106) for MDR1 (A) and BCRP1 (B) in total MNCs (blasts) from 28 AML samples are plotted against the MFIRs for the same samples (determined by calculating mitoxantrone-induced fluorescence in the presence of cyclosporine A divided by fluorescence in the absence of cyclosporine A as described in the Materials and methods section). Correlation coefficients (r) and P-values (t test) are indicated.

Figure 4.

The MFIR in newly diagnosed AML patients varies with response to induction chemotherapy. MFIRs of total PB MNCs (blasts) and CD34+CD38 cells from 14 CR (•) and 14 chemotherapy refractory (NR ▪) newly diagnosed AML patients. Values for individual patient samples and the means (horizontal lines) are shown for each group. P-values (t test) comparing CR and NR patient samples are shown.

Table 1. AML Patients Characteristics
  Total (n = 174)Induction failure (n = 47)Complete remission (n = 127)P = IF vs. CR
  • a

    Mann Whitney U test.

  • b

    Chi square.

  • c

    Prognostic groupings according to MRC (UK) criteria (39).

  • d

    Fms-like tyrosine kinase-3 internal tandem duplication.

  • e

    e?>Induction chemotherapy; 7 + 3 = cytarabine 200 mg/m2/day for 7 days + daunorubicin 45 mg/m2/day for 3 days; HiDAC/DNR = cytarabine 1.5 g/m2 bid for 6 days + daunorubicin 45 mg/m2/day for 3 days; INDAC = cytarabine 1.5 g/m2/day for 3 days; MITOX = mitoxantrone 12 mg/m2/day for 3 days; VP16 = etoposide 800 mg/m2 for one dose.

Age (years)Number (% of total)55 (17–78)59 (17–78)56 (18–75)0.31a
Age ≥ 60 77 (44)23 (49)54 (43)0.45b
Cytogeneticsc number (% of total)Favorable16 (9)0 (0)16 (13)0.005b
Intermediate131 (75)34 (72)97 (76)
Unfavorable18 (10)8 (17)10 (8)
Normal80 (46)19 (40)61 (48)0.37b
Unknown/failed9 (5)5 (11)4 (3) 
FLT3 ITDd+/− (% +)36/138 (21)15 (32)21 (17)0.03b
WBC (×109/L) at diagnosisMedian (range)30.6 (0.5–431)37.2 (0.8–431)26.7 (0.5–376)0.26a
Induction chemotherapye?> number (% of total)7 + 377 (44)29 (62)48 (38)0.02b
HiDAC /DNR63 (36)13 (28)50 (39)
HiDAC or INDAC + MITOX + VP1624 (14)5 (10)19 (15)
Other10 (6)0 (0)10 (8)
Table 2. Comparison of MFI Ratio Determined from Marrow or Peripheral Blood (A) or Fresh vs. Cryopreserved Cells (B)
Patient #MFI ratio
  1. (A) The MFIR was determined from total MNCs in paired cryopreserved peripheral blood (PB) and bone marrow (BM) samples from 12 newly diagnosed AML samples. These were samples selected from the 174 described on Table 1.

  2. (B) Fresh PB and BM samples from three patients were divided into two aliquots, and the MFIR was measured immediately from one aliquot (fresh) while the second aliquot was cryopreserved, left frozen for several days, and then thawed for analysis (frozen). These three samples are not part of the cohort of 174 described on Table 1.

 Mean ± SD2.0 ± 0.72.0 ± 0.7

Mononuclear cells (MNCs) were isolated by Ficoll-Hypaque density gradient centrifugation (Pharmacia, Uppsala, Sweden) and cryopreserved in Iscove's modified Dulbecco's medium (IMDM) with 50% fetal calf serum (FCS) (StemCell Technologies, Vancouver, Canada) and 10% dimethylsulfoxide (Sigma-Aldrich, Oakville, Ontario, Canada). Thawed cells were washed twice in IMDM containing 10% FCS and used for the experiments described below.

Induction Chemotherapy

Induction chemotherapy was cytarabine and anthracycline-based for all patients. Chemotherapy for the 174 AML patients whose clinical data was used for the logistic regression analysis is described in Table 1. Seventy-seven received one or two courses of conventional dose cytarabine (200 mg/m2/day for 7 days) with daunorubicin (DNR) 45 mg/m2 for 3 days (7+3), 63 patients received cytarabine 1.5 g/m2 bid for 6 days (HiDAC) + DNR 45 mg/m2/day for 3 days. Twenty-four patients received mitoxantrone 12 mg/m2/day for 3 days in combination with cytarabine 1.5 g/m2/day (INDAC) or 3.0 g/m2/day for 3 days and etoposide 800 mg/m2 for one dose. Ten patients received other therapy including 7+3 with mylotarg (n = 1), 7+3 with PSC833 (n = 3), CPX351 (a liposomal formulation of cytarabine and DNR (n = 1), HiDAC alone (n = 2) or in combination with DNR and fludarabine (n = 1), or in combination with DNR and etoposide (n = 1), or in combination with DNR, vincristine, and prednisone (n = 1).

CR was defined as described by Cheson et al. (40, 41). One hundred twenty-seven (73%) of patients achieved CR1. Twenty-two of these required two cycles of chemotherapy to achieve CR; in nine cases, the additional chemotherapy was 7+3, in five high dose cytarabine with DNR and in eight cases 2.4 g/m2 etoposide with cytoxan 2 g/m2 × 3 (42). Forty-seven (27%) of the patients failed to achieve CR1. In 29 cases, IF was determined after one or two courses of 7+3 chemotherapy, while the remaining 18 IF patients received high dose cytarabine with DNR or mitoxantrone/etoposide as the first cycle of induction therapy followed by salvage chemotherapy with 2.4 g/m2 etoposide with cytoxan 2 g/m2 × 3 (42) (n = 14), a second course of high dose cytarabine with an anthracycline (n = 2) or gemtuzumab ozogamicin (n = 2). Overall survival (OS) was defined as the interval from diagnosis to death independently of the cause.

QRT-PCR Analysis

Total RNA was extracted from 1 × 106 unsorted AML blast cells, CD34+CD38+, and CD34+CD38 AML cells with Absolutely RNA Miniprep, Microprep, or Nanoprep kits (Stratagene, La Jolla, CA). The RT reaction was performed in 20 μL with superscript III reverse transcriptase (Invitrogen, Burlington, Canada) using random hexamer oligonucleotides (Amersham Pharmacia, Piscataway, NJ). Real-time PCR was performed using 12.5 μL SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA), 1 μL of 20 pM-specific primers, 1 to 2 μL cDNA, and water to a final volume of 25 μL. Specific forward and reverse primers to produce ∼ 100-bp amplicons for optimal amplification in real-time PCR of reverse-transcribed cDNA for human MDR1 were 5′-GGCCTAATGCCGAACACATT-3′ (forward) and 5′-AGGCTCAGTCCCTGAAGCAC-3′ (reverse), for human BCRP1 were 5′-CCAGGCGAAGGTTTTTCACA-3′ (forward) and 5′-TGGGACTGGTTATAGGTGCCA-3′ (reverse), and human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were 5′-CCCATCACCATCTTCCAGGAG-3′ (forward) and 5′-CTTCTCCATGGTGGTGAAGACG-3′(reverse). Thermal cycling conditions were 50°C for 2 min and 95°C for 5 min, followed by 40 cycles of 15 s at 95°C, 30 s at 58°C, and 30 s at 72°C. Real-time PCR and data analysis were performed on an iCycler iQ system, using iCycler iQ Real-time Detection Software (Bio-Rad, Hercules, CA). Fold-expression relative to GAPDH was calculated by: (1 + AE)dCt, where AE is the amplification efficiency of the specific gene and dCt = (Ct of Gene X) − (Ct of GAPDH).

Mitoxantrone Efflux for Assessment of ABC Transporter Function

ABC transporter function was assessed with flow cytometry using mitoxantrone (Sigma-Aldrich) as a fluorescent substrate and cyclosporine A (CA) (Sigma-Aldrich) as an inhibitor of ABC transporter function. ABC transporter-mediated mitoxantrone efflux was quantified by determining the MFI of AML cells incubated with this drug alone, when compared with the MFI measured in cells incubated with mitoxantrone plus CA (43, 44). MNCs isolated from patients' PB or BM were washed with Hanks buffered solution (HBS) with 2% FCS and 0.02% sodium azide (HFN) and titrated to a cell count of 5 × 105/mL in prewarmed IMDM media with 10% FCS. Mitoxantrone (0.1 μg/mL) was added with or without CA (10 μM), and cells were incubated for 1 h at 37°C. Subsequently, cells were washed and incubated for an additional 1 h in mitoxantrone-free 10% FCS containing IMDM with or without CA to allow time for mitoxantrone efflux. In some experiments, cells were then washed in HBS and stained with anti-CD34-PE and anti-CD38-FITC for 30 min at 4°C (Becton Dickinson, Franklin Lakes, NJ) to allow analysis of subpopulations of cells enriched for progenitor cells or anti-CD3-APC-Cy7 and CD19-APC-Cy7 (Becton Dickinson) to allow exclusion of B and T lymphocytes. Fluorescence-activated cell sorting (FACS) analysis was performed with a FACScalibur flow cytometer (Becton Dickinson) (excitation wavelength of 635 nm and emission wavelength of 661 nm). Figure 1 shows a representative example of the FACS analysis performed. MFI was calculated using Flowjo software (Tree Star, Ashland, OR) for cells stained with mitoxantrone with or without CA after machine calibration using unstained control cells. The MFI in the presence of CA divided by the MFI in the absence of CA was calculated for each AML sample and cell population and reported as the MFIR in the results reported below. In experiments where cells were stained with anti-CD34/CD38 or anti-CD3/CD19, mouse IgG antibodies conjugated with each fluorochrome were used as isotype controls. Gates were set to exclude nonviable, propidium iodide positive (PI+) cells, and cells labeled with the irrelevant isotype control antibody. The MFI for cells incubated with mitoxantrone ± CA was calculated for subpopulations defined by their expression of CD34 and CD38 using a gating strategy previously described (45) and for cells negative for CD3 and CD19 by comparison with isotype controls (Fig. 1).


The correlation between mRNA expression level of ABC transporters and between ABC transporters and MFIR for the same patient sample was estimated using Spearman's rank-order correlation coefficient. Comparison of mean MFIR values between patient groups was performed using the Student t test. Comparison of categorical variables and medians was performed using chi square and Mann Whitney U tests, respectively. The predictive value of MFIR with respect to CR rate was determined using logistic regression analysis. Clinical parameters independently associated with response were determined in multivariate logistic regression analysis. P values ≤ 0.05 were considered statistically significant.


MDR1 and BCRP1 Expression in Leukemic Blasts and Subpopulations from AML Patients Varies with Response to Induction Chemotherapy

Previous work has shown that expression of the major ABC transporters, MDR1 and BCRP1 is higher in AML blasts from patients who fail to achieve CR with remission induction chemotherapy (NR patients) than in blasts from patients who successfully enter CR, particularly when the CD34+CD38 subpopulation enriched for candidate LSC is analyzed (37). In this study, a similar analysis was performed using QRT-PCR to study MDR1 and BCRP1 expression in total blasts, CD34+CD38+, and CD34+CD38 subpopulations from 14 CR and 14 NR AML patients. As shown in Figure 2, the expression of both MDR1 and BCRP1 was on average higher (four-fold) in NR than CR patient samples whether the CD34+CD38 subpopulation or total blasts were evaluated (P < 0.04 for all comparisons). There was a significant correlation between MDR1 and BCRP1 expression in the same AML sample (r = 0.7, P < 0.01). Thus, high expression of two ABC transporters in leukemic blasts at diagnosis predicted clinical drug resistance in these AML patients.

Assessment of ABC Transporter Function by Flow Cytometry

To develop a flow cytometry-based assay to quantify ABC transporter function, mitoxantrone was chosen as a fluorescent ABC transporter substrate and cyclosporine (CA) as a nonspecific inhibitor of ABC transporter protein function, including that of MDR1 and BCRP1. As described in the Materials and methods section and shown in Figure 1, the MFI of AML cells incubated with both mitoxantrone and cyclosporine is divided by the MFI of cells incubated with mitoxantrone alone to create a ratio (MFIR).

Comparison of the MFIR calculated for total blasts from the same 28 AML patient samples that were analyzed by QRT-PCR (Fig. 3) revealed a direct correlation between the MFIR and the QRT-PCR results for both MDR1 (r = 0.78, P < 0.0001) and BCRP1 (r = 0.63, P < 0.0004). Comparison of the MFIR values from total blasts and CD34+CD38 cells between NR and CR patient samples revealed significant differences with the mean MFIR in the NR group being substantially higher (mean ± SD; 2.18 ± 0.55 and 2.39 ± 0.59 for total blasts and CD34+CD38 cells, respectively) than the same mean MFIR values for CR patients (1.44 ± 0.40 and 1.64 ± 0.47, respectively) (P < 0.001 for both comparisons between NR and CR patients; Fig. 4).

Thus, among this small group of AML patient samples calculation of the MFIR seemed to have similar predictive value for chemotherapy responsiveness as QRT-PCR for MDR1 and/or BCRP1.

Comparison MFIR between PB and BM, and between Fresh and Frozen Samples

The above analysis was performed using cryopreserved PB samples from patients with high circulating blast counts. To determine if blasts from the PB and BM would give similar results, paired PB and BM samples from 12 patients were tested. As shown in Table 2, MFIRs of PB and BM blasts were very similar for all patients (mean ± SD, 2.0 ± 0.7 and 2.0 ± 0.7, respectively). Similarly, as described in previous studies, when the MFIR from cryopreserved and fresh PB and BM cells were compared for three patient samples, no significant difference was observed (Table 2; 44,46).

MFIR Association with Response to Induction Chemotherapy in 174 AML Patients

To attempt to confirm the value of the MFIR assay for predicting the outcome of induction chemotherapy and to further analyze its possible prognostic use for independently predicting this and other outcomes, a larger group (n = 174) of AML patient samples was analyzed, which is distinct from the initial 28 samples described above (for which data is shown in Figs. 2, 3, 4, 2–4). To minimize the risk of investigator bias treatment, outcome was unknown at the time the MFIR was determined for these 174 samples.

The characteristics of AML patients used for this analysis are listed in Table 1. Among 174 patients who received remission induction therapy, 105 (60%) achieved CR after one cycle of induction therapy (CR1), 22 (13%) required two cycles to achieve CR (CR2), and 47 (27%) were refractory to induction therapy (IF). As expected, the cytogenetic risk profile (39) varied between CR and IF patients with all patients in the good risk group achieving CR, while a larger proportion of unfavorable risk patients were in the IF group. Similarly, a larger proportion of IF than CR patients showed the FLT3 internal tandem duplication (FLT3 ITD). A larger proportion of CR patients received high or intermediate dose cytarabine-containing induction regimens rather than conventional 7+3, when compared with the IF patients. The relationship between CR status and MFIR in total PB MNCs from these newly diagnosed patients is shown in Table 3. In addition, to determine if the predictive value of the test could be improved by focusing on the subpopulation enriched for LSC or MNC from which T and B lymphocytes had been depleted, CD34+CD38 and CD3CD19 cells from some of these same samples were also analyzed. The mean MFIR for the CR patients was significantly lower than that obtained for IF patients and the mean MFIR for patients obtaining CR after two cycles of chemotherapy was higher than that obtained for those obtaining CR after one cycle when total MNCs, CD34+CD38, or CD3CD19 cells were compared (P < 0.001 for all comparisons). Among the three cell populations studied, the difference between CR and IF samples was greatest for CD3CD19 cells. The mean MFIRs for CD34+CD38 cells were, on average, higher for all patients, regardless of their remission status. However, restricting analysis to this cell population did not enhance the ability to discriminate CR from IF patients.

Table 3. The MFIR is Lower in AML Patients Achieving CR with One Cycle of Chemotherapy than in Those Requiring Two Cycles to Enter CR or Those Refractory to Induction Chemotherapy
 MFIR mean ± SD
Total MNCa (N = 174)CD34+CD38 (N = 174)CD3CD19 (N = 71)
  • a

    Total PB mononuclear cells (blasts).

CR (1 cycle)1.62 ± 0.531.96 ± 0.801.62 ± 0.50
CR (2 cycles)2.22 ± 1.292.21 ± 1.202.99 ± 1.47
Induction failure2.59 ± 0.983.01 ± 1.303.15 ± 0.83
P < (t test)0.0010.0010.001

On univariate analysis, the MFIR did not vary with cytogenetic risk group, presenting WBC count or the presence of the FLT3 ITD but was higher in patients above 60 years, when compared with younger patients (see below).

Determination of a Threshold MFIR Value Associated with IF

To determine an MFIR value above which IF might be predicted, a logistic regression analysis was performed using data from the 174 patients from Table 1. This analysis determined 2.45 as the MFIR threshold above which achievement of CR (after one or two cycles of chemotherapy) was unlikely. The logistic regression also determined that the MFIR determined using either total MNCs or CD3CD19 MNCs allowed a more clear separation of CR and IF patients than that determined using CD34+CD38 cells. Therefore, only the results obtained with the former two analyses are shown. Although the overall CR rate for the 174 patients was 73%, 117 (84%) of the 140 patients with MFIR ≤ 2.45 calculated from total MNC achieved CR, when compared with only 10 (29%) of the 34 patients with MFIR > 2.45 (P < 0.0001) (Table 4).

Table 4. The MFIR Predicts Induction Failure in 174 AML Patients
 Induction failure (n =)Complete remission (n =)% CRP = (chi square)
  1. Total PB MNC MFIR.

MFIR > 2.45241029< 0.0001
MFIR ≤ 2.452311784 
All patients4712773 

MFIR Association with Response to Induction Chemotherapy Varies between CR and IF Patients Receiving Either Conventional or High Dose Cytarabine in Combination with DNR

To assess the predictive value of the MFIR in groups of patients who had all received the same induction chemotherapy and to determine if the use of high dose (HiDAC) vs. conventional dose cytarabine would change this predictive value, the subgroups from the original group of 174 AML patients receiving 7+3 induction (n = 77) or HiDAC with DNR (n = 63) were analyzed separately. As shown in Table 5, patients receiving 7+3 were on average older and had less favorable cytogenetics than those receiving HiDAC + DNR. Nevertheless, the MFIR was higher in IF than in CR patients in both treatment groups (P ≤ 0.004) (Table 5). Patients whose AML blasts showed an MFIR > 2.45 showed a much lower CR rate (after one or two cycles of chemotherapy) than those with MFIR ≤ 2.45 in both chemotherapy groups 26% vs. 78% and 20% vs. 84% for 7+3 and HiDAC + DNR, respectively, P ≤ 0.0003 for both comparisons (Table 6).

Table 5. The MFIR Varies Significantly Among Complete Remission and Induction Failure Patients Receiving Daunorubicin in Combination with Either Conventional (7 + 3) or High Dose Cytarabine (HiDAC)
  7 + 3 (n = 77)HiDAC + DNR (n = 63)P =
  • a

    Chi Square test.

  • b

    Mann Whitney U test.

  • c

    Student t test. NS; P > 0.3.

  • 7 + 3 = cytarabine 200 mg/m2/day for 7 days + daunorubicin 45 mg/m2/day for 3 days; HiDAC + DNR = cytarabine 1.5 g/m2 bid for 6 days + daunorubicin 45 mg/m2/day for 3 days.

Age (years)median (range)63 (19–78)44 (17–59)< 0.001b
Cytogenetics number (% of total)Favorable3 (4)10 (16)0.016a
Intermediate66 (86)40 (63.5) 
Unfavorable6 (8)11 (17.5) 
Unknown/failed2 (2)2 (3) 
Normal41 (53)24 (38)0.07a
FLT3 ITD+/− (% +)19/58 (25)9/54 (14)0.13a
WBC (×109/L) at diagnosisMedian (range)23.8 (0.8–431)24.5 (0.5–306)NSb
MFIR (mean ± SD)CR (1 cycle)1.74 ± 0.58 (n = 40)1.62 ± 0.49 (n = 39) 
CR (2 cycles)2.18 ± 0.74 (n = 8)1.70 ± 0.38 (n = 11) 
Induction failure2.71 ± 0.91 (n = 29)2.18 ± 0.63 (n = 13) 
P = (comparing mean MFIRs between CR (1 or 2 cycles) and induction failurec <0.00010.004 
Table 6. A High MFIR Predicts Induction Failure Among Patients Treated with Either Conventional (7 + 3) or High Dose Cytarabine (HiDAC) in Combination with Daunorubicin
 Induction failure (n =)Complete remission (n =)% CRP = (chi square)
  1. Total PB MNC MFIR.

(A) 7 + 3 (n = 77)
MFIR > 2.4517626<0.0001
MFIR ≤ 2.45124278
All patients294862 
(B) HiDAC + Daunorubicin (n = 63)
MFIR > 2.4541200.0003
MFIR ≤ 2.4594984
All patients135079 

MFIR Values Associated with IF in NK AML

Cytogenetic abnormalities detected in diagnostic BM samples are the most powerful prognostic factor in newly diagnosed AML. However, at least 40% of AML patients are present with a NK and are classified as having intermediate risk disease (7). To determine if MFIR analysis would retain its predictive value in NK AML, a subgroup analysis was performed using data from the 80 NK patients. Although this group received a variety of chemotherapy induction regimens, these did not vary significantly between CR and IF patients. As shown in Table 7, six (33%) of 18 patients with an MFIR > 2.45 obtained CR (after one or two cycles of chemotherapy), when compared with 55 (89%) of 62 patients with MFIR ≤ 2.45 (P < 0.0001) demonstrating that this measurement in total MNC retained its association with chemotherapy response in the NK group.

Table 7. The MFIR Predicts Induction Failure in 80 Normal Karyotype AML Patients
 Induction failure (n =)Complete remission (n =)% CRP = (chi square)
  1. MFIR determined from total PB MNC. The median (range) age and presenting WBC count for these 80 patients were 56 years (18–75) and 65.7 × 109/L (0.5–376), respectively, with no significant difference between the induction failure (IF) and complete remission (CR) patients (P = 0.91 and 0.1, respectively, Mann Whitney U test). Induction chemotherapy consisted of 7 + 3, HiDAC/DNR, mitoxantrone-containing regimens or other regimens in 41, 24, 11, and 4 patients, respectively. There was no significant difference between IF and CR patients in the proportion receiving these different regiments (P = 0.46, chi square). The FLT3 ITD was present in 11 of 19 IF and 15 of 61 CR patients (P = 0.01, chi square).

MFIR > 2.4512633<0.0001
MFIR ≤ 2.4575589
All patients196161 

MFIR Values Associated with IF in Elderly AML Patients

Increasing age has a negative prognostic impact in AML with lower overall remission rates and higher treatment related complications in the elderly (12). Consistent with the relative chemotherapy resistance seen in older AML patients, in this group of 174 AML patients, the mean total MNC MFIR was higher among patients above 60 years than among younger patients (mean ± SD 2.17 ± 1.04 and 1.78 ± 0.74 for older and younger patients, respectively, P < 0.004). The ability to predict the outcome of remission induction would be particularly relevant in elderly AML patients. Thus, further analysis was performed restricted to the 77 patients greater than age 60 at diagnosis of AML. Among 54 patients with MFIR ≤ 2.45, 47 (87%) achieved CR (after one or two cycles of chemotherapy), while CR was obtained for only seven (30%) of 23 patients with higher MFIR values (P < 0.0001, Table 8), indicating that the association of the MFIR with chemotherapy response is retained in this elderly patient group.

Table 8. The MFIR Predicts Induction Failure Among 77 AML Patients ≥ 60 years old at Diagnosis
 Induction failure (n =)Complete remission (n =)% CRP = (chi square)
  1. MFIR determined from total PB MNC. The median (range) age and presenting WBC count for these 77 patients were 66 years (60–78) and 56.0 × 109/L (0.8–328), respectively, with no significant difference between induction failure (IF) and complete remission (CR) patients for these characteristics (P = 0.58 and 0.46, respectively, Mann Whitney test). The cytogenetic risk group was good, intermediate, poor or unknown for 2, 65, 5, and 5 patients, respectively, while 42 patients had a normal karyotype and 21 had the FLT3 ITD. There was no significant difference between IF and CR patients for these characteristics (P = 0.33, 0.44 and 0.48, respectively, chi square). Induction chemotherapy consisted of 7 + 3 for 59 (77%) of patients while 13 (12 in the CR group) received a mitoxantrone-containing regimen and five (all in the CR group) received other therapy. A larger proportion of CR than IF patients received a mitoxantrone-containing or other regimens (P = 0.03, chi square).

MFIR > 2.4516730< 0.0001
MFIR ≤ 2.4574787
All patients235470 

Further Refinement of the MFIR by Restricting Analysis to CD3CD19 Cells

Because the analysis of mean MFIR values shown in Table 3 and the logistic regression analysis had suggested that restricting analysis to CD3CD19 cells might enhance the predictive power of this test, samples from 38 NK and 70 elderly AML patients where testing had been done on CD3CD19 cells were analyzed separately. Table 9 shows the CR rate of 93% in NK AML patients with MFIR ≤ 2.45, which was much higher than the 20% CR rate seen with patients whose MFIR was above this threshold (P < 0.0001). Similar results were seen among the 70 elderly AML patients where those with the low MFIR had a higher probability of achieving CR, when compared with the high MFIR group (90% vs. 30%, P < 0.0001, Table 9).

Table 9. MFIR Determined on CD3CD19 AML PB Cells Predicts Outcome in NK and Elderly AML Patients
 Induction failure (n =)Complete remission (n =)% CRP = (chi square)
(A) Normal karyotype (n = 38)
MFIR > 2.458220<0.0001
MFIR ≤ 2.4522693
All patients102874 
(B) Patients ≥ 60 years (n = 70)
MFIR > 2.4514630<0.0001
MFIR ≤ 2.4554590
All patients195173 

MFIR Varies with OS but not CR Duration

Additional logistic regression analysis was performed to determine if the MFIR could predict either CR duration or OS. No predictive value for this measurement was seen for CR duration greater than or less than 1 year among the 126 CR patients (P = 0.15). However, an MFIR value ≤ 1.91 was able to predict patients who would achieve survival of ≥ 1 year (the median survival for the entire group of 174 patients; P < 0.0001, Table 10).

Table 10. The MFIR Predicts Overall Survival (OS) ≥ or < 1 year
 OS < 1 yrOS ≥ 1 yrTotalP = (chi square)
N = (% of total)N = (% of total)
  1. Logistic regression analysis was used to determine the MFIR value (from total MNC) that was most predictive of OS from diagnosis ≥ 1 year.

MFIR > 1.9144 (70)19 (30)63<0.0001
MFIR ≤ 1.9143 (39)68 (61)111 
All patients87 (50)87 (50)174 

A High MFIR Associated Independently with IF

Finally, to determine if the MFIR determination was an independent predictor of the probability of achieving CR with induction chemotherapy a stepwise multiple logistic regression analysis was performed. In this analysis, the variables considered in addition to the MFIR were cytogenetics (good, intermediate, or poor risk by MRC (UK) criteria), age at diagnosis (≥ or < 60 years), presenting white blood cell count (≥ or < the median of 30 × 109/L), the presence or absence of FLT3 ITD and chemotherapy regimen. Only the MFIR value determined using either total MNC or CD3CD19 cells showed independent predictive value in this model (P < 0.0001 for both MFIR determinations, P ≥ 0.4 for all other variables).


Although several ABC transporter genes (ABCA2, ABCA3, ABCB1, ABCB4, ABCB5, ABCB11, ABCC1∼6, and ABCG2) are associated with chemotherapeutic drug resistance, MDR1 and BCRP1 are the principal multidrug resistance (MDR) proteins that have been identified in AML blasts including leukemia stem cells (LSCs) (47–51). In our own previous work QRT-PCR was used to measure expression of all 49 ABC transporters thus far identified in man in AML blasts from newly diagnosed patients (37). However, only MDR1 and BCRP1 expression in AML cells could be clearly associated with chemotherapy refractoriness (37). Although this previous work clearly implicates ABC transporter function in the MDR phenotype so often seen in AML the use of such data to enhance assessment of prognosis has not been widely adopted, likely at least partly due to the difficulty with clinical implementation and standardization of the techniques used.

In previous studies, a variety of assays have been used to determine ABC transporter expression including QRT-PCR and flow cytometry using monoclonal antibodies specific for individual transporters (37, 52–54). In this analysis, flow cytometry was used to assess ABC transporter function. Mitoxantrone was selected as the ABC transporter substrate since its efflux from AML cells can be mediated by various transporters including both MDR1 and BCRP1. In addition, our early comparison of mitoxantrone with DNR and other reports suggested assays using the former drug would be more sensitive for measurement of drug efflux (43, 55). Cyclosporine A was chosen as the transporter inhibitor as it inhibits the function of both MDR1 and BCRP1. Thus, the combined use of mitoxantrone and cyclosporine A allows the assay to give a more global assessment of ABC transporter function than measurement of individual gene products. Consistent with this assumption, Figure 3 shows the good correlation observed between the MFIR efflux assay and QRT-PCR for MDR1 and BCRP1. The MFIR assay requires small numbers of cells from clinical PB or BM samples that are routinely available from newly diagnosed AML patients and technology (flow cytometry), which is widely available and frequently used in clinical hematology laboratories.

The goal of developing this assay was to produce a test that had strong predictive value for assessing the likely outcome of induction chemotherapy. As shown by Tables 3–9, this preliminary analysis of a large group of unselected newly diagnosed AML patients suggests that the MFIR assay might meet these criteria. Importantly, when the analysis was restricted to patients who had received homogeneous chemotherapy regimens, NK AML or patients above 60 years, the predictive value was retained. The latter two patient groups are examples of those where additional prognostic information would be particularly useful. Most of the samples analyzed in this retrospective study were cryopreserved PB cells from patients with relatively high circulating blast counts. In clinical practice, it is expected that many AML patients would present with few circulating leukemia cells or a mixture of blasts and normal hematopoietic elements in PB and that fresh rather than cryopreserved samples would be analyzed. Thus, a series of experiments were done to compare MFIR values obtained from PB and BM or fresh and cryopreserved cells from the same patient. As shown in Table 2, no significant difference was seen between PB and BM samples suggesting that blasts from either source will give comparable results in the MFIR assay. Similarly, as reported by others, cryopreservation did not change the MFIR value obtained demonstrating that fresh samples routinely available for clinical diagnostics could be successfully used for testing (44, 46).

Because of concerns raised by previous reports suggesting that contaminating lymphocytes could increase the apparent ABC transporter expression in AML PB (56) and to ensure that the majority of PB cells used to calculate the MFIR were AML blasts, further experiments were done in which the MFIR was determined after gating out CD3+ and CD19+ T- and B-cells. As shown in Table 9, the MFIR retained its predictive value when CD3CD19 cells were analyzed. Further analysis of larger numbers of samples with low peripheral blast counts will be necessary to determine if this refinement to the MFIR assay substantially improves its predictive value.

Considerable data has accumulated demonstrating that malignant hematopoiesis in AML is organized as a hierarchy with a primitive subpopulation of leukemia-initiating or “stem” cells, which maintain the leukemic clone (35, 57). These candidate LSCs display properties of relative chemotherapy resistance including cell cycle quiescence, over expression of ABC transporters, and intrinsic mitoxantrone and DNR efflux capacity (36, 37, 58). Thus, LSC likely contribute to failure of induction chemotherapy and relapse after CR is obtained. In many (but not all) AML samples, the CD34+CD38 subpopulation of blasts is greatly enriched for LSC (35). Thus, experiments were done to determine if the MFIR assessed in these cells would have enhanced value for predicting chemotherapy outcomes. Figure 2 shows that, consistent with our previous report, QRT-PCR values for MDR1 and BCRP1 seemed to show somewhat greater difference between CR and NR patient samples when the analysis was restricted to CD34+CD38 cells, when compared with total blasts or CD34+CD38+ cells. Nevertheless, significant differences between chemotherapy responsive and refractory patient samples were seen for all three cell populations. This difference between total blasts and CD34+CD38 cells was less apparent when the MFIR was calculated for these populations (Fig. 4, Table 3). This suggests that LSCs transmit at least some of their MDR phenotype to their progeny and that the MFIR assessed on total blast cells can be a good predictor of outcome to initial chemotherapy.

The logistic regression analysis was extended to determine if the MFIR could predict other treatment outcomes in addition to achievement of CR. As would be expected from the effect of the MFIR on CR rates, a significant association between high MFIR values and short OS was seen (Table 10). However, among 126 CR patients the MFIR did not predict remission duration. This result is consistent with the likelihood that relapse is mediated by subclones of drug resistant leukemia cells that emerge after the large majority of chemotherapy sensitive cells have been eliminated. Although we did not have the opportunity to measure the MFIR at relapse in this retrospective study, prior reports have documented increased ABC transporter expression in blasts from relapsed AML patients (59–61).

Finally, to determine if the MFIR provided value for predicting the outcome of induction chemotherapy that was independent of other established prognostic variables in AML, a stepwise multiple logistic regression analysis was performed. In this model which included cytogenetic risk group (39), age, WBC count, and FLT3 ITD as well as the MFIR as variables, only the MFIR determined using either total MNC or CD3CD19 PB cells retained statistical significance. Although this result may seem surprising, given the large number of studies validating the listed prognostic variables in AML, it is likely explained by the specific endpoint that was evaluated (achievement of initial CR rather than overall or disease free survival) and the relatively small number of patients in this study. In any case, the data suggest that the MFIR as determined in this study might ultimately be validated as a strong predictor of the likelihood of successful induction chemotherapy.

Thus, we believe that this rapid assessment of ABC transporter function is potentially a relevant assay to include in the work-up of newly diagnosed AML patients. The MFIR test could be particularly useful for selecting therapy for patients where the toxicity of induction chemotherapy is predictably high; e.g., AML patients who are elderly or who have severe comorbid illness and may also be helpful for patients where other prognostic markers are either unavailable or unhelpful, e.g., NK AML. Further evaluation of this assay is clearly needed before its clinical value can be determined. To further this aim, a prospective evaluation of this MFIR testing using clinical laboratory services and a larger cohort of AML patients is planned.