|Home | About | Journals | Submit | Contact Us | Français|
In a patient with acute myeloid leukemia (AML) following therapy, finding ≥5% bone marrow (BM) blasts is highly concerning for residual/relapsed disease. Over an 18 month period, we performed multicolor flow cytometry immunophenotyping (MFC) for AML minimal residual disease (MRD) on >4,000 BM samples, and identified 41 patients who had ≥5% myeloblasts by morphology but no leukemic blasts by MFC. At the time of a negative MFC, an abnormal cytogenetic study converted to negative in 14 patients and remained positive at a low level (2.5 to 9.5%) by fluorescence in situ hybridization in 3 (14%), of the latter, the abnormalities subsequently disappeared in the repeated BM in 2 patients. Positive pre-treatment mutations, including FLT3, NPM1, IDH1, CEBPA, became negative in all 10 patients tested. Of the 7 patients with favorable cytogenetics, PML/RARA, CBFB-MYH11 or RUNX1-RUNX1T1 fusion transcripts were detected at various levels in 6 patients but all patients remained in complete remission in the end of follow-up. With no additional chemotherapy given, 39 patients had BM repeated (median 2 weeks, range <1 to 21), and all cases showed <5% BM blasts and a continuously negative MFC. In the end of follow-up (median 10 months, range 1–22), 13 patients experienced relapse, 12/13 showing clonal cytogenetic evolution/switch and 11 demonstrating major immunophenotypic shifts. We conclude that MFC is useful in identifying a regenerating BM sample with >5% BM blasts that would otherwise be scored as positive using standard morphologic examination. We believe this conclusion is supported by the changes in molecular cytogenetic status and the patient clinical follow-up data.
Acute myeloid leukemia (AML) is a biologically heterogeneous disease characterized by clonal accumulation of immature myeloid cells . With the current treatment strategies, ~60–80% AML patients achieve complete remission (CR); however, most patients will eventually relapse . Persistence of minimal residual leukemic cells is thought to be an important cause of relapse; and there is a growing body of evidence that detection of subclinical levels of leukemia (i.e. minimal residual disease, MRD) provides independent prognostic information [3–6]. The current criteria for CR however, are still defined as hematological recovery and <5% BM blasts by microscopic blast enumeration . Conversely, finding ≥5% BM blasts by morphology is considered as positive for residual or relapsed AML. It is known, however, that there is substantial morphological similarity between leukemic blasts and regenerating hematopoietic precursors, and that morphological assessment has a low sensitivity and specificity. Furthermore, outcome data have shown the need for a more objective and robust measure of CR than morphologic BM blast count [7–9].
The most commonly used methods to determine MRD are multicolor flow cytometry (MFC) and molecular-genetic techniques. Molecular MRD offers very high sensitivity for detection of fusion transcripts (up to 10−4 to 10−5) but can only be used in less than one-third of patients. Molecular MRD detection by other techniques such as next-generation sequencing for recurrent point mutations may be applicable in more cases, but is typically less sensitive (10−2 to 10−3). Multicolor flow cytometry offers good sensitivity (up to 10−4) and is applicable in >90% of all AML cases [10–12]. A positive MRD has been shown to correlate with leukemic event free survival and overall survival in patients post induction, consolidation as well as to predict transplant outcomes[3, 5, 13, 14]. In our laboratory, we have developed and validated an AML MRD MFC assay which combines identification of a leukemia associated immunophenotype (LAIP)” and detection of “deviation from normal” approaches . Persistence of AML MRD at 30 and 90 days by our MFC assay after achieving complete remission was shown to be the most powerful predictors for patient outcome.
In this study, we focus on a group of AML patients who after therapy had a BM sample showing ≥5% BM blasts by morphology, but was tested negative by MFC analysis. By correlating with molecular and genetic data and follow-up information, we investigated the reliability of MFC in the presence of an overwhelming number of regenerating myeloid precursors, and sought to better understand the clinical significance of these discrepant results in adult AML patients.
From January, 2013 to June, 2014 we performed AML MRD MFC assays on 4,043 BM samples, which included BM samples collected post induction, in the course of maintenance therapy, surveillance and post hematopoietic stem cell transplant (HSCT). Cases with a negative MFC finding but a BM blast count ≥5% were identified and included in this study. The clinicopathologic information was obtained by review of the medical charts. This study was approved by the Institutional Review Board of MDACC.
Morphologic evaluation was performed independently without knowledge of the MFC or molecular genetic findings. Each case was reviewed at least by one hematopathologist, and the diagnosis of cases that fulfilled the initial inclusion criteria was confirmed by another hematopathologist for this study. For each case, routine hematoxylin and eosin-stained (H&E) histologic sections of BM biopsy and aspirate clot, and Wright–Giemsa-stained BM aspirate smears were evaluated. A 500 cell count was performed based on examination of multiple fields of BM aspirate smears. Cases with inadequate quality of BM smears were not included in this study.
BM aspirate samples were collected in EDTA-anticoagulant and processed within 24 hr of collection. After incubation with monoclonal antibodies for 10 min at 4°C, erythrocytes were lysed with ammonium chloride (PharmLyse™, BD Biosciences, San Diego, CA) at room temperature for 10 min using a standard lyse/wash technique. The antibody panel is shown in Supplemental Table 1. Samples were acquired on FACSCanto II instruments (BD Biosciences, San Diego, CA, USA) which were standardized daily using CS&T beads. A minimum of 200,000 live events were acquired to achieve a potential sensitivity of at least 10−4 (0.01%). Instrument alignments, sensitivities, and spectral compensation were verified by standards, calibrators, procedural controls and normal peripheral blood samples prior to processing of patient samples.
In each tube, after excluding debris and doublets, a CD45 dim ‘blast’ gate including monocytes and a CD34+ cell gate was drawn (Figure 1). Both the CD34+ myeloblasts and CD34− populations of the CD45 dim+ cells were separately evaluated, following the analytic approach published previously [15, 17]. In brief, AML leukemic blasts differed from normal myeloid precursors in several categories: 1) altered expression levels of antigens normally expressed, either decreased (CD38, HLA-DR) or increased (CD13, CD34, CD117, CD123); altered CD45/side scatter pattern, 2). Asynchronous expression of myelomonocytic antigens on myeloblasts (CD4, CD64, CD36, CD15), and/or 3). Aberrant expression of lymphoid antigens, (CD2, CD5, CD7, CD19, CD22, CD56). For normally expressed antigens, the levels of expression were measured by mean fluorescence intensity (MFI); and for antigens not normally or only partially expressed, expression was measured as a percentage of blasts. The AML MRD validation study was initially performed in 30 BM lymphoma stating BM prior to chemotherapy. The normal ranges were established on 100 AML patients post HSCT, and proven to be in clinical, morphological, molecular (polymorphism study) and cytogenetic remission since these BM sample showed more regenerating changes. These HSCT samples also helped us to correlate MFI with visual inspection. For example, CD38-APC had a MFI of 28,822, and a SD = 4752. A 1/3 log scale MFI change was equivalent to at least 1 standard deviation, and mostly 2 standard deviation difference. This visual inspection for each parameter was less subject to sample quality and instrument variation but remained reproducible among three hematopathologists “Alterations” were two standard deviations from normal, also confirmed by at least 1/3 of log scale changes by visual inspection of the data. Critical markers (CD13, CD34, CD38, CD45, CD123, CD117) were primarily assessed by MFI, and confirmed by visual inspection. From the initial validation study of AML MRD assay (n=67), immunophenotypical aberrancy was identified essentially in all AML patients at diagnosis, although in some cases (5–10%), the changes were considered to be less pronounced or only involving 1 marker.
The detection of ≥20 clustered events with two or more aberrant markers, whether present in the original AML or not, was considered to be positive for MRD. In certain instances, a positive diagnosis might still be rendered with only one aberrant marker if antigen expression was markedly deviated from normal or was part of the LAIP. The same was true if fewer than 20 events showed multiple, highly distinctive aberrations. The latter situation, a minimum of 10 dots was required. MRD levels were reported as a percentage of total events.
Conventional chromosomal analysis was performed on G-banded metaphase cells prepared from unstimulated 24- and 48-hr BM aspirate cultures at the time of diagnosis using standard techniques. The median number of metaphases analyzed was 20 (range: 10–50). The karyotype was documented according to the International System for Human Cytogenetic Nomenclature(ISCN 2013) .
Interphase fluorescence in situ hybridization (FISH) analysis was performed in cases with known abnormalities by conventional chromosomal analysis at follow-up according to the manufacturer’s instructions (Abbott Molecular, Abbott Park, IL). Combined morphologic and FISH analysis was performed on a subset of cases with the methods described previously  with minor modifications. In brief, morphologic evaluation and image capture were performed on BM aspirate smears with Wright-Giemsa stain (x100); smears were then destained using 1% acid alcohol, followed by protease II (Abbott Molecular) treatment and hybridization with probes. Two hundred nuclei were counted and percentage of abnormal cells was calculated. The target cell populations were captured under fluorescent microscope.
Nanofluidics-based qualitative multiparametric reverse-transcriptase PCR was performed for recurrent fusion transcripts including: t(8;21)(q22;q22); RUNX1-RUNX1T, inv(16)(p13.1q22) or t(16;16)(p13.1q22); CBFB-MYH11 variant A and CBFB-MYH11 variant D, variant E; t(15;17)(q22;q12); PML-RARA long form, short form and alternative form, t(9;22)(q34;q11.2); BCR-ABL1 b2a2, b3a2, e1a2, and t(6;9)(p23;q34); DEK-NUP214. For cases positive for any fusion product, real-time reverse transcription polymerase chain reaction (PCR) was routinely performed for monitoring treatment response . Values were expressed as a percentage of normal ABL transcripts. Mutation analysis was performed by direct sequencing followed by capillary gel electrophoresis using ABI Prism 3100 Genetic Analyzer (Applied Biosystems). In some cases, mutation analysis was performed by next generation sequencing (NGS) using a customized TruSeq Amplicon Cancer Panel and a MiSeq sequencer (illumina Inc, San Diego, CA). Library preparation and sequencing using MiSeq were performed according to the manufacturer’s instructions . The genes tested in the panels included DNMT3A, EZH2, FLT3, GATA1, GATA2, HRAS, IDH1, IDH2, IKZF2, JAK2, KIT, KRAS, MDM2, MLL, MPL, NOTCH1, NPM1, NRAS, PTPN11, RUNX1, TP53, and WT1; and in a subset of cases also TET2, ASXL1, and NOTCH genes.
Correlations were evaluated using Pearson’s correlation coefficient. The statistical analyses were performed using SPSS software (IBM Corporation, Armonk, NY). Results were considered statistically significant if p-values were less than 0.05 in a two-tailed test.
Over an 18 month period we performed AML MRD MFC analysis on over 4,000 BM samples at our hospital. As shown in Supplemental Figure 1, 1177 of 1236 (95%) samples (646 patients) with ≥5% BM blasts detected by morphology were also positive by MFC MRD study; 15 (1%) samples were indeterminate by MFC MRD. In 44 (4%) samples (44 patients) with BM blasts ≥5% (5–19%), MFC yielded a negative result. Of these 44 patients, three were excluded (two cases evolved to pure erythroid leukemia and one to histiocytic sarcoma) and MFC analysis did not obtain neoplastic cells for analysis.
The remaining 41 cases formed this study group. Thirty-two (32) patients were treated for AML newly diagnosed AML and 9 patients were treated for relapsed AML. By the treatment time course, 27 patients were status post induction chemotherapy; 8 patients were in the course of consolidation/maintenance chemotherapy, and 6 were post HSCT. Only one of the patients received G-CSF within one week of the BM examination. The median time from last chemotherapy to MFC study was 3.5 weeks (1.0 week to 6.5 weeks). For 6 post HSCT patients, the median time from HSCT to MFC study was 5 weeks (4–17 weeks). According to the criteria of the National Cancer Institute-sponsored workshop on AML , 26 (63%) patients had insufficient platelet recovery (<100 × 109/L) and 30 (73%) patients had insufficient neutrophil recovery (ANC <1×109/L) at the time of analysis. The demographic and clinicopathologic features and the AML WHO classification of these 41 patients are shown in Table 1.
Multicolor flow cytometry immunophenotyping showed findings consistent with a normal regenerating BM in all patients, and there was no evidence of AML detected by either by LAIP or “deviation from normal” assessment. A case pre-treatment and post-treatment MFC and morphological findings are shown in Figure 1. The median level of CD34+ myeloid precursors detected by MFC was 2.0% (0.1–10.0%), the number was not proportional to the blast number counted morphologically (median 7%; 5–19%, r=0.015, p=0.927). Of 41 patients, the AML blasts in the pre-treatment BM were CD34+ in 20 cases, CD34 partial+ in 8 patients; and CD34- in 13 cases. Seven cases were myelomonocytic; and 4 cases were monocytic.
At diagnosis, conventional cytogenetic analysis was performed in 40 patients and FISH was performed in 20 patients. Twenty-two of 40 (55%) patients had an abnormal karyotype, including 7 cases with a favorable karyotype (Table 2), 3 with a complex karyotype and 12 with other abnormalities.
At the time when BM showed ≥5% blasts but MFC MRD study was negative, a total of 37 patient BM samples were submitted for chromosomal analysis and 9 also for FISH analysis. Thirty-four (34) patients had a normal cytogenetic result. Three patients had a normal karyotype; however, FISH revealed persistent abnormality in a small number of interphases, including 1 case with 9.5% monosomy 7; 1 with 2.5% PML-RARA; and 1 case with 7.0% monosomy 7 cells (laboratory cutoff is 5%).
Combined morphology-FISH analysis using probes D7S522 (7q31, spectrum orange) and CEP7 (centromere, spectrum green) (Abbott Molecular, Inc) was performed in two cases with monosomy 7. In the case with 7% of cells positive for monosomy 7, abnormal signals were observed in maturing/differentiated forms but not in blasts (Figure 2); whereas, in the case with 9.5% monosomy 7 cells, the abnormal signals were observed in both the mature cells and blasts. In the next follow-up BM samples, the case with 7% monosomy 7 (2 weeks later) and the case with 2.5% PML-RARA (1 week later), FISH became negative. However, in the patient with 9.5% monosomy 7, monosomy 7 was again detected at 6.5% interphases at follow-up BM performed 8 weeks later after 1 course of consolidation chemotherapy. AML eventually relapsed, showing clonal cytogenetic evolution (Supplemental Table 2, case 1).
Molecular studies were performed at the time of AML diagnosis in 39 patients, of which 23 were done by NGS. Various molecular abnormalities were detected in 27 patients, including 7 patients with recurrent cytogenetic abnormalities. At the time of a negative MFC study, one case with CBFB-MYH11 became negative, two showed a low level of PML/RARA fusion transcripts (0.02% and 0.04%), two had low levels of CBFB-MYH11 (0.06% and 2.22%) and two had RUNX1-RUNX1T1 (1.64%, and 87.57%). Ten patients had previously detected mutations and a paired sample tested at the time of MFC study, including 7 with FLT3 ITD, 1 with NPM1, 1 with NPM1 and IDH1 and one CEBPA. All mutations became non-detectable at the time of a negative result of MFC.
None of these patients received additional dose of chemotherapy for the reported BM blast count ≥5% because of a negative MFC study. Two patients died of transplant-related complications prior to have a repeated BM examination. In the remaining 39 alive patients, BM examination was repeated in a median of 2 weeks (5 days to 21 weeks). In the repeated BM biopsy, all 39 patients showed <5% BM blasts, and negative MFC results.
The median follow-up length was 10 months (range: 1–22) for all patients, including alive and dead. There were a median of 4 (range: 0–10) subsequent BM examinations and 3 (range: 0–9) MFC analyses performed. Over the follow-up period, 13 patients experienced AML relapse and 25 remained in continuous CR or CR with incomplete blood count recovery (CRi), 1 patient lost follow-up and 2 died of HSCT related complications. Twelve of 13 (92%) relapsed patients showed clonal cytogenetic evolution or clonal switch (supplemental Table 2). NRAS/KRAS mutations, initially detected in 3 patients, were no longer detectable at relapse. NGS studies were performed in 5 patients at diagnosis and at relapse, mutations in NPM1, ASXL1, TET2, NOTCH, IDH1/2, EZH2 and DNMT3A were again detected and notably, no new mutations were detected at relapse (supplemental Table 2).
It is noteworthy that in the 6 patients who had a positive RT-PCR result for recurrent fusion transcripts, PML/RARA became negative whereas 2 patients with low levels of CBFB-MYH11 and 2 patients with RUNX1-RUNX1T1 remained positive in the follow-up BM samples. All 6 patients were in continuous CR/CRi at the end of the follow-up (median follow-up 14.5 months).
In this study, we investigated 41 AML cases that BM showed ≥5% blasts at various points in the treatment course of AML, but were negative by MFC MRD analysis. By correlating with molecular cytogenetic data, follow-up BM findings and patient outcome, we show that MFC designed for AML MRD detection can clearly differentiate a “regenerating marrow” from residual leukemic blasts in a morphologically “positive” BM.
A number of studies have shown that morphologic blast counts may not correlate well with MFC MRD results. In a multicenter study of children and adolescents with acute leukemia , 8.2% BM samples with <5% blasts were MRD positive by MFC, whereas 57.5% BM samples with ≥5% blasts were negative by MRD. Similarly, Loken et al.  reported in a Children’s Oncology Group study that 25% of patients who met morphologic criteria for CR were positive by MFC MRD and that 26% of patients who failed to achieve morphologic CR (blasts reported as 5–20%) had no detectable MRD by MFC. Both studies showed that MFC results are prognostically relevant and that morphological assessment is of limited value in assessment of treatment effect in pediatric AML. In our study, the frequency of patients with ≥5% BM blasts but a negative MFC study was less than 5% by sample numbers. Notably, our patients were almost exclusively adults, and many were older than 60 years. Furthermore, many of the study patients tested had high risk AML that included secondary AML, therapy-related AML, AML with high risk cytogenetics, and refractory/relapsed AML. In fact, we detected a significantly higher positive MRD rate by MFC in our patient population, either in BM with ≥5 or <5% blasts (ASH abstract 1015, 2014). As a matter of fact, although the information of treatment status and AML subtype was not collected in all 4043 BM samples, the proportion of cases with favorable cytogenetics (17%) in our study group appeared to be higher than that of our AML patient population in the same period of time (8%). Therefore, a lower discordant rate in our patients with ≥5% BM blasts was likely due to a combination of an older patient age who had less robust BM regenerating capability and a high proportion of high risk AML that were less likely to achieve true CR. On the other hand, of the 44 cases we initially identified, two were AML with progression to pure erythroid leukemia (PEL) and one to histiocytic sarcoma. In these cases, MFC, either by LAIP or “deviation from normal”, failed to detect leukemic blasts. These three cases illustrate the limitations of MFC in such variants of AML and highlight the importance of morphological and MFC correlation in AML assessment.
In the analysis of the 41 cases included in this study, 7 cases were AML with recurrent cytogenetic abnormalities, including 2 t(15;17); 3 inv(16) and 2 t(8;21). Six of 7 patients were positive for fusion transcripts by RT-PCR at various levels. In 2 patients with acute promyelocytic leukemia, PML/RARA fusion transcripts converted to negative in the follow-up in post-therapy BM; whereas, in two patients with CBFB-MYH11 and two with RUNX1-RUNX1T1 RT-PCR assays remained positive at a low level at the end of the follow-up even though these patients were in continuous clinical CR. Studies have shown that RUNX1-RUNX1T1 transcripts may remain detectable in patients in CR, even many years after therapy; and in RUNX1-RUNX1T1+ AML long-term survivors may even harbor relatively high number of copies of the fusion product [22–25]. Similarly, for CBFB-MYH11 AML, about 20–30% of RT-PCR-negative patients eventually relapsed and few long-term survivors achieved complete PCR negativity [24, 26, 27]. Perea et al. found that in the early phases of CR of patients with t(8;21) and inv(16), there were no differences in fusion copy numbers between relapsed and non-relapsed patients. In contrast, detection of >1% MRD by MFC in the end of induction was prognostically significant . Conceivably, low levels of MRD by PCR could be controlled by subsequent chemotherapy or immune reconstitution and hence fail to predict relapse. It is recommended that MRD assays by MFC may help these groups of patients to avoid unnecessary additional induction chemotherapy .
Cytogenetic abnormalities were detected in 3 patients by FISH when MFC MRD assays were negative. In 2 patients, the low levels of abnormalities detected by FISH were no longer detectable in the follow-up BM, but in 1 patient, abnormality persisted. Interestingly, by the combined morphology-FISH study, we showed that if cytogenetic abnormalities were seen in the mature cells rather than blasts, they were likely to convert to negativity by FISH with longer follow up; whereas, if an abnormal signal was identified in blasts, it was likely to be persistent. These findings suggest that in some cases, cytogenetic abnormalities are in partially differentiated leukemic cells that have lost their clonogenic potential. On the other hand, it also indicates that the number of leukemic blasts may be below the level of detection of the MFC assay, especially in cases with fewer LAIPs and the presence of overwhelming numbers of normal regenerating myeloid precursors. Nevertheless, FISH analysis for previously identified cytogenetic abnormalities is complementary to MFC MRD analysis.
Although no MRD was detected by MFC, about 30% patients experienced relapse with a median of 10 month follow-up. A well-control study to compare survival and relapse with AML MRD positive cases was difficult in considering the heterogeneity of AML subtypes and treatment status in these patients. However, relapses in our patients appeared to be less than a 50–70% relapse rate in patients with morphological remission but a positive MRD at 30 days and 90 days post induction, and more close to a 20–30% relapse rate in AML MRD-negative patients in our previous study. In our 13 clonal cytogenetic evolution and/or regression with new emerging clones were identified in 12 patients. NRAS or KRAS mutations were detected in 3 patients at the time of initial diagnosis, but become negative at relapse. This finding indicates that KRAS/NRAS mutations are likely to be a secondary event that may involve a subclone of the AML cells rather than a tumor initiating event; and leukemic cells harboring these mutations may be sensitive to high dose chemotherapy. On the other hand, mutations including ASXL1, TET2, IDH1/2, NPM1, NOTCH1 and EZH1 were unchanged at relapse. We also observed a major phenotypic switch at relapse (in 11/13 patients). These findings underscore the complexity of AML at the biological level, which is particularly true in adult AML, secondary AML and relapsed AML. Genetic evolution of AML has been shown to be a dynamic process shaped by multiple cycles of mutation acquisition and clonal selection [30–32]; and clonal evolution or clonal selection are likely to occur during disease treatment [33–35]. DNMT3A mutations, for example, have been found in patients in clinical remission at various allele frequencies, likely in preleukaemic stem cells that are resistant to chemotherapy. These data may explain why a negative MRD does not portend a sustainable remission in some patients. On the other hand, heterogeneity of leukemic blasts imposes a diagnostic challenge for AML MRD detection [37, 38]. Our findings also show the importance of incorporating a “deviation from normal” approach in addition to detecting LAIP in MFC MRD analysis . This combined approach not only ensures the detection of all populations that appear to be “abnormal” at diagnosis, but also will be less subjective to tumor heterogeneity, phenotypic drift and clonal selection during the course of treatment.
In conclusion, recent studies have shown multicolor flow cytometry immunophenotyping (MFC) for minimal residual disease (MRD) is highly useful for detecting evidence of persistent AML after therapy when BM morphologic examination is negative. In this study, we show that MFC is also helpful in the uncommon circumstance when BM examination shows ≥ 5% BM blasts and yet MFC is negative. All 39 alive patients had <5% blasts by morphology and a continuously negative MFC on the repeated BM. These patients did not receive any therapy between the two BM evaluations. This suggests that MFC clearly identifies “regenerating marrow” as compared to morphology. Our correlation with cytogenetic molecular data and clinical follow up supports the view that compared to morphologic examination, MFC is a more reliable means of assessing for AML residual disease. Our findings indicate that reliance on pure morphology not only results in a significant number with an incomplete response to therapy being withheld further treatment, but also, a significant number of patients being incorrectly deemed refractory and receiving salvage chemotherapy when this may not be warranted.
Conflict of interest: all authors have no conflict of interest to declare