Pathology (December 2015) 47(7), pp. 609–621
REVIEW
The role of multiparametric flow cytometry in the detection of minimal residual disease in acute leukaemia DENISE LEE1,2, GEORGE GRIGORIADIS3,4,5,6
AND
DAVID WESTERMAN1,2
1Peter MacCallum Cancer Centre, 2University of Melbourne, 3Department of Clinical Haematology, Monash and Alfred Health, 4Alfred Pathology Service, 5Southern Clinical School, Monash University, and 6Centre for Cancer Research, MIMR-PHI of Medical Research, Melbourne, Vic, Australia
Summary Flow cytometry is the most accessible method for minimal residual disease (MRD) detection due to its availability in most haematological centres. Using a precise combination of different antibodies, immunophenotypic detection of MRD in acute leukaemia can be performed by identifying abnormal combinations or expressions of antigens on malignant cells at diagnosis, during and post treatment. These abnormal phenotypes, referred to as leukaemia-associated immunophenotypes (LAIPs) are either absent or expressed at low frequency in normal bone marrow (BM) cells and are used to monitor the behaviour and quantitate the amount of residual disease following treatment. In paediatric acute lymphoblastic leukaemia (ALL), the level of MRD by multiparametric flow cytometry (MPFC) during therapy is recognised as an important predictor of outcome. Although less extensively studied, adult ALL and adult and paediatric acute myeloid leukaemia (AML) have also demonstrated similar findings. The challenge now is incorporating this information for riskstratification so that therapy can be tailored individually and ultimately improve outcome while also limiting treatmentrelated toxicity. In this review we will elaborate on the current and future role of MPFC in MRD in acute leukaemia while also addressing its limitations. Key words: Acute lymphoblastic leukaemia, acute myeloid leukaemia, minimal residual disease. Received 13 November 2014, revised 10 August, accepted 14 August 2015
INTRODUCTION Minimal residual disease (MRD) refers to residual leukaemic cells in the setting of morphological remission (90% of ALL. Several studies have demonstrated that MRD monitoring of fusion genes (PML-RARA, RUNX1-RUNX1T1, CBFB-MYH11) by realtime quantitative PCR (RT-qPCR) provides prognostic information17–19 and monitoring of PML-RARA in acute promyelocytic leukaemia (APL) is well-established for guiding therapy.20,21 Mutated NPM1 is present in 30% of all AML cases and has proven to be a clinically significant MRD marker, predictive of relapse and survival.22–24 Next-generation sequencing (NGS) technology provides an alternative platform for molecular MRD detection, enabling a high-throughput of samples while also abrogating many of the limitations of MPFC and RT-PCR. Salipante et al.25 observed that the sensitivity of detection for NPM1 by NGS methods was 0.001%, more than a log greater than for MPFC, but comparable to RT-PCR. NGS also eliminated the requirement for multiple allele-specific probes or prior knowledge of the NPM1 mutation subtype. While there is no doubt molecular MRD, particularly in the context of NGS technology, is an informative marker of relapse, a detailed discussion on the comparison of methodologies for MRD assessment in acute leukaemia is beyond the scope of this review; we refer the readers to Grimwade and Freeman26 and Bruggemann et al.27 Our review will focus on
2015 Royal College of Pathologists of Australasia. All rights reserved.
Copyright © Royal College of pathologists of Australasia. Unauthorized reproduction of this article is prohibited.
610
Pathology (2015), 47(7), December
LEE et al.
the current and future role of MPFC in MRD in acute leukaemia while also addressing its limitations.
METHODOLOGIES IN MRD ANALYSIS IN AML Leukaemia associated immunophenotypes (LAIPs) A LAIP is a unique and specific combination of antigens identified on leukaemic cells at diagnosis, making them ideal for sequential MRD monitoring. It should consist of a progenitor antigen (CD34, CD117, HLA-DR for AML; CD34, TdT for ALL), a lineage-specific antigen (CD3/19 for T/B cells respectively and MPO for myeloid cells) and aberrantly expressed antigen(s). Further qualitative information can be gained from aberrant physical characteristics with side- and forward-scatter properties.28 The LAIP may be present on all or a subset of leukaemic cells and should be absent or present only in low frequency on normal bone marrow cells.10 Multiple LAIPs10,29 are often identified and all should be followed during treatment to maximise specificity and accuracy. Figure 1 is an example of MRD analysis in AML. Up to 0.5% of normal bone marrow cells may express a LAIP, compromising specificity29 and resulting in inaccurate quantification of MRD, particularly at lower levels of detection (90% of AML patients with sensitivities of detection ranging from 102 to 104. They also demonstrated that patients with post-induction levels of hMICL/CD123 above the median in regenerating BM had a shorter RFS ( p < 0.001). The evolution of a ‘universal’ MRD marker is growing, with a similar approach being investigated by the EuroFlow Consortium to use single-tube disease-oriented panels for MRD detection over patient-specific antibody combinations.39 This concept of a standardised MRD marker panel will be pivotal in establishing consistency in MRD analysis. ‘Different from normal’ The ‘different from normal’ approach to MRD analysis overcomes the problem of MRD false negatives due to immunophenotypic shifts. Identification of MRD is reliant on comparing patterns of normal antigenic expression through all stages of maturation against the sample of interest. Populations are considered abnormal if they are 0.5 decades away from the position of the corresponding normal cells.8,40 Another advantage to this approach compared to LAIP identification is that MRD analysis is still possible despite not having an available diagnostic LAIP, which may occur when samples are sent to a centralised laboratory for evaluation. This method of analysis necessitates expert knowledge of normal haematopoietic maturation patterns, however, which may be a major limiting factor in its widespread implementation in laboratories. Achieving standardisation in MRD analysis when applying this approach is also more complex, as the determination of ‘normal’ patterns can be somewhat subjective. Each laboratory should establish normal expression with their antibody panel using a minimum of 20 patient/donor samples. We recommend both methods be routinely used when performing MRD analysis. However, this approach significantly increases the time spent per evaluation, therefore workflow and efficiency requirements of diagnostic laboratories should be considered before a laboratory offers MRD testing.
STANDARDISATION Definition of a LAIP One of the key difficulties in standardising MPFC-based MRD monitoring stems from a lack of consensus regarding the true definition of a LAIP. Different groups adopt conflicting criteria resulting in variably reported LAIPs. According to Feller et al.10 and Terwijn et al.41 a LAIP was considered valid only if expressed on at least 10% of leukaemic blasts. Alternatively, Buccisano et al.42 defined a LAIP to be expressed on at
Copyright © Royal College of pathologists of Australasia. Unauthorized reproduction of this article is prohibited.
FLOW CYTOMETRY IN ACUTE LEUKAEMIA Diagnostic AML
[Singlets]
103
[WBCs]
103
[CD45 progenitor gate]
103
CD117+
Viable cells
102
101
102 CD117 PC5
CD45 V500-A
102 SSC-A
611
101 CD45 progenitor gate
100
100
0
0
101 2 0
200
400 600 FSC-A [CD117+]
800
0
1000
10
1
10 SSC-A [CD11b-]
2
10
10
3
102
103
102
102
102
1 102 101 HLA-DR APC-H7
103
CD33 V450
103
5
101 SSC-A
[CD13+ dim]
103
101
100
0
103
CD13 PE
CD11b APC
0
–2 0
101
101 CD13+ dim
2
100
0 0
0.2 0
–5 –2 –4 –2
0
2
4
101 CD34 PE-Cy7
102
–4 –2
103
0
2
4 101 CD34 PE-Cy7
102
–1
103
0
MRD AML [WBCs]
[Singlets]
[CD45 progenitor gate]
103
103
103
CD117+
Viable cells 2
10
101
100
CD117 PC5
102 CD45 V500
SSC-A
102
101
101
CD45 progenitor gate
100
2
0 –0.5
0
0 0
0
200
400 600 FSC-A
800
1
0
10 SSC-A
10
1000 Gate All
2
10
10
3
–2
Number
[CD11b-] [CD117+] 10
101 5
101
101 100
100
MRD
0
–2
0
101 2 CD34 PE-Cy7
102
103
0
CD13+ DIM
0.2 0
–5
103
102 CD33 V450
CD13 PE
CD11b APC
CD11B-
102
103
102
102
101 SSC-A
[CD13+ dim] HLA-DR APC-H7 / CD33 V450
103
3
100
0
390,895
–1 –2
0
101 2 CD34 PE-Cy7
102
0
1
101 102 HLA-DR APC-47
103
103 Gate number All 1,152 MRD 1,152
Fig. 1 Diagnostic AML: Initial sequential gating strategy to include viable cells only (not shown): event number vs time, FSC-A vs FSC-H and FSC-A vs SSC-A. LAIP comprises abnormal antigen expression compared to normal blasts (CD34, dim CD13þ, partial HLA-DR). MRD AML: A sequential gating strategy is used in combination with physical characteristics (SSC-A vs FSC-A) to identify minimal residual disease (MRD). MRD is discriminated from promyelocytes (shown in purple) by their physical characteristics, high side scatter, HLA-DR positivity and bright CD13 expression, compared with dim CD13 expression on the leukaemic blasts. Normal blast expression (arrow) is confirmed by comparing to blasts from a healthy control. MRD ¼ 0.29% (MRD events/total WBCs 100). Black, leukaemic myeloblasts/MRD; purple, promyelocytes. WBC, white blood cells.
Copyright © Royal College of pathologists of Australasia. Unauthorized reproduction of this article is prohibited.
612
LEE et al.
least 50% of blasts. Different still, Olaru et al.43 deemed a LAIP to be significant if it was expressed on at least 1% of the total population (comprising 100,000 CD45 positive events). Denominator Discrepancy amongst authors also exists regarding the denominator used in MRD calculations. The most common denominator used is the total white blood cell (WBC) count. The lysing reagent eliminates the majority of red cells during sample preparation, (although this process can be imperfect) before further exclusion by a CD45 vs side scatter gating strategy. However, in paediatric ALL the denominator is mononuclear cells, which is based on the tradition of using ficoll to extract the mononuclear component. Although both techniques were recognised as sample preparation methods at the Proceedings of the Second International Symposium on MRD assessment,44 the recently published British Committee for Standards in Haematology (BCSH) Guidelines recommend red cell lysis over density centrifugation.45 Number of events to acquire The total number of events recommended for acquisition ranges from 100,00043 to 250,0007,46 to 1 106.41,44 Variability amongst groups regarding what constitutes an ‘event’ also influences sensitivity with some groups aiming for total events, which would include debris and erythroid precursors, and others aiming for total WBC events. Both of these methods are open to errors in enumeration due to peripheral blood contamination in bone marrow aspirate samples.47 Attaining adequate event detection can also be problematic in paucicellular marrows. Techniques to address these limitations include doubling the standard volume of cells, density gradient centrifugation or ‘bulk lysis’.48 Sample quality can also be assessed by ensuring adequate levels of cells that are normally restricted to the marrow such as erythroid, myeloid and B-cell progenitors are present, a technique recommended by the European Myeloma Network when performing myeloma MRD.49 The number of events that defines a leukaemic population also varies significantly between diseases and groups. The IBFM-ALL-FLOW MRD Network recommends a minimum of 30 leukaemic events per tube accompanied by a confirmatory, independent tube.44 The EuroFlow Consortium recommends at least 100 leukaemic cells (as the sum of events in all tubes) in line with their recommendation for MPFC MRD quantitation of multiple myeloma.49 In our laboratory we aim to acquire >500,000–1,000,000 events in AML and ALL. The appropriate absolute cut-off of abnormal leukaemic cells varies on an individual patient basis, but 50–100 abnormal events would usually ensure a sensitivity around 1 in 10,000. Reproducibility The perceived lack of reproducibility of MPFC, particularly across multicentre studies, has so far hindered the progress of standardisation.39,50 Reproducibility is influenced by a large number of complex factors including equipment and reagent variability and stability, software, and data analysis and interpretation. At the heart is a thorough and robust quality framework. The International Council for Standardisation of Haematology (ICSH), the International Clinical Cytometry Society (ICCS) and the BCSH have released independent
Pathology (2015), 47(7), December
practice guidelines in an attempt to address the issues of standardisation and promote inter-laboratory harmonisation.45,51 The recommendations provide guidance on instrument set up, fluorochrome and panel design, sample preparation, data analysis and quality control. It is hoped these guidelines will improve uniformity of practice across laboratories for all diagnostic cases. Notably, MRD analysis was not addressed in any of the aforementioned guidelines. The EuroFlow Consortium published recommendations for MRD markers in antibody panels, but also did not specifically address standardisation of MRD analysis. Feller et al.52 demonstrated that standardisation of LAIPs for MRD analysis is possible in a multicentre setting. This was achieved by defining a standardised antibody panel and standard operating procedures. Newer developments such as automated analysis algorithms may also be helpful in improving uniformity.53
ACUTE MYELOID LEUKAEMIA Prognostic relevance of MRD following induction and consolidation San Miguel et al.9 were the first to demonstrate the prognostic significance of flow-based MRD in AML, identifying two levels of MRD (0.5% and 0.2%, respectively) that correlated with a higher risk of relapse post-induction and post-intensification. In a later study by the same group, the authors showed that post-induction MRD levels could also stratify patients into different risk groups.11 Buccisano et al.42 demonstrated that post-consolidation MRD levels >0.035% correlated with an unfavourable prognosis in terms of relapse rate, OS and RFS ( p < 0.001). In contrast to San Miguel et al.,11 they found that post-induction MRD status bore no influence on outcome, provided MRD negativity was achieved by the end of consolidation.42 Postconsolidation MRD levels proved to be an independent predictor of outcome when compared against karyotype, age and post-induction MRD levels as covariates. Effect of chemotherapy on MRD The type of chemotherapy administered may affect the level of MRD achievable and re-calculation may be necessary for treatment protocols that involve more intensive regimens resulting in greater tumour debulking and potentially less MRD.4,9,42 The findings of Rubnitz et al.54 and Feller et al.10 would contradict this suggestion where treatment regimens did not appear to have any impact on the degree of MRD positivity. Timing The optimal time-point for MRD assessment is still unresolved with some advocating post-induction,8,11,41 post-consolidation,6,42 or both.7,10 Timing is critical when applying riskadapted therapy. Premature evaluation would potentially identify more patients resulting in additional (and unnecessary) intensive therapy. However, deferred evaluation (i.e., postconsolidation) may have limited practical utility. While there is no consensus regarding the optimal time-point for MRD assessment, earlier evaluation (post-induction) is likely to be the most useful so that treatment could be altered if required.1,5,8,11,41,54,55
Copyright © Royal College of pathologists of Australasia. Unauthorized reproduction of this article is prohibited.
FLOW CYTOMETRY IN ACUTE LEUKAEMIA
Also unclear is the frequency of MRD assessments. Although regular monitoring may provide peace of mind, this would involve frequent and possibly unnecessary bone marrow biopsies, which are not without risks. Intimately associated with this question is whether earlier therapy (i.e., in the absence of morphological relapse) ultimately improves survival. We look forward to prospective, clinical trials that may adequately answer these questions. Quality of response The kinetics of blast reduction has been found to be independently associated with OS. The German AML Cooperative Group determined that morphologically detectable blasts midinduction indicated slow clearance and was a negative predictor of event-free survival (EFS), OS and RFS.56 This has been reaffirmed in later trials demonstrating early blast clearance as an independent predictor of survival post-induction.57,58 Maurillo et al.59 disagreed with these findings, suggesting that achievement of MRD negativity by the end of consolidation is more important than achieving early MRD negativity. They demonstrated that 30% of patients who were MRD positive post induction went on to achieve MRD negativity following consolidation, with no significant difference in RFS and OS between MRD negative patients post-induction and consolidation and MRD positive patients post-induction who converted to MRD negative status following consolidation. This would be concordant with the findings of Buccisano et al.42 and Venditti et al.,6 suggesting the kinetics of blast clearance may not be an accurate predictor of survival, which seems counter-intuitive. Alternatively, Rubnitz et al.54 suggested that the depth of response might be more predictive of outcome. Here, a riskadaptive approach incorporating serial MRD assessments (following induction 1 and 2) was applied to guide therapy intensification in a paediatric AML cohort.54 The outcome for patients with low levels of MRD (0.1% to 0.1% post-induction 1 and 2 was an independent predictor of OS, compared with cytogenetics and white cell
count (WCC) ( p ¼ 0.037, HR 3.79, and p ¼ 0.022, HR 6.15, respectively). These findings and their approach to MRD assessment subsequently informed the risk assignment algorithm in the multicentre AML2002 study, which was discussed earlier.54 Sievers et al.8 demonstrated that patients with MRD >0.5% after remission induction were 4.8 times more likely to relapse ( p < 0.0001) with a mortality rate that was 3.1 times greater ( p < 0.0001). Following consolidation, OS at 3 years was 41% vs 69% for patients with >0.5% vs 30 109/L, t(4;11) or other MLL rearrangement) were assessed for MRD by 4-colour flow cytometry to guide treatment stratification. Patients with MRD levels
REVIEW
The role of multiparametric flow cytometry in the detection of minimal residual disease in acute leukaemia DENISE LEE1,2, GEORGE GRIGORIADIS3,4,5,6
AND
DAVID WESTERMAN1,2
1Peter MacCallum Cancer Centre, 2University of Melbourne, 3Department of Clinical Haematology, Monash and Alfred Health, 4Alfred Pathology Service, 5Southern Clinical School, Monash University, and 6Centre for Cancer Research, MIMR-PHI of Medical Research, Melbourne, Vic, Australia
Summary Flow cytometry is the most accessible method for minimal residual disease (MRD) detection due to its availability in most haematological centres. Using a precise combination of different antibodies, immunophenotypic detection of MRD in acute leukaemia can be performed by identifying abnormal combinations or expressions of antigens on malignant cells at diagnosis, during and post treatment. These abnormal phenotypes, referred to as leukaemia-associated immunophenotypes (LAIPs) are either absent or expressed at low frequency in normal bone marrow (BM) cells and are used to monitor the behaviour and quantitate the amount of residual disease following treatment. In paediatric acute lymphoblastic leukaemia (ALL), the level of MRD by multiparametric flow cytometry (MPFC) during therapy is recognised as an important predictor of outcome. Although less extensively studied, adult ALL and adult and paediatric acute myeloid leukaemia (AML) have also demonstrated similar findings. The challenge now is incorporating this information for riskstratification so that therapy can be tailored individually and ultimately improve outcome while also limiting treatmentrelated toxicity. In this review we will elaborate on the current and future role of MPFC in MRD in acute leukaemia while also addressing its limitations. Key words: Acute lymphoblastic leukaemia, acute myeloid leukaemia, minimal residual disease. Received 13 November 2014, revised 10 August, accepted 14 August 2015
INTRODUCTION Minimal residual disease (MRD) refers to residual leukaemic cells in the setting of morphological remission (90% of ALL. Several studies have demonstrated that MRD monitoring of fusion genes (PML-RARA, RUNX1-RUNX1T1, CBFB-MYH11) by realtime quantitative PCR (RT-qPCR) provides prognostic information17–19 and monitoring of PML-RARA in acute promyelocytic leukaemia (APL) is well-established for guiding therapy.20,21 Mutated NPM1 is present in 30% of all AML cases and has proven to be a clinically significant MRD marker, predictive of relapse and survival.22–24 Next-generation sequencing (NGS) technology provides an alternative platform for molecular MRD detection, enabling a high-throughput of samples while also abrogating many of the limitations of MPFC and RT-PCR. Salipante et al.25 observed that the sensitivity of detection for NPM1 by NGS methods was 0.001%, more than a log greater than for MPFC, but comparable to RT-PCR. NGS also eliminated the requirement for multiple allele-specific probes or prior knowledge of the NPM1 mutation subtype. While there is no doubt molecular MRD, particularly in the context of NGS technology, is an informative marker of relapse, a detailed discussion on the comparison of methodologies for MRD assessment in acute leukaemia is beyond the scope of this review; we refer the readers to Grimwade and Freeman26 and Bruggemann et al.27 Our review will focus on
2015 Royal College of Pathologists of Australasia. All rights reserved.
Copyright © Royal College of pathologists of Australasia. Unauthorized reproduction of this article is prohibited.
610
Pathology (2015), 47(7), December
LEE et al.
the current and future role of MPFC in MRD in acute leukaemia while also addressing its limitations.
METHODOLOGIES IN MRD ANALYSIS IN AML Leukaemia associated immunophenotypes (LAIPs) A LAIP is a unique and specific combination of antigens identified on leukaemic cells at diagnosis, making them ideal for sequential MRD monitoring. It should consist of a progenitor antigen (CD34, CD117, HLA-DR for AML; CD34, TdT for ALL), a lineage-specific antigen (CD3/19 for T/B cells respectively and MPO for myeloid cells) and aberrantly expressed antigen(s). Further qualitative information can be gained from aberrant physical characteristics with side- and forward-scatter properties.28 The LAIP may be present on all or a subset of leukaemic cells and should be absent or present only in low frequency on normal bone marrow cells.10 Multiple LAIPs10,29 are often identified and all should be followed during treatment to maximise specificity and accuracy. Figure 1 is an example of MRD analysis in AML. Up to 0.5% of normal bone marrow cells may express a LAIP, compromising specificity29 and resulting in inaccurate quantification of MRD, particularly at lower levels of detection (90% of AML patients with sensitivities of detection ranging from 102 to 104. They also demonstrated that patients with post-induction levels of hMICL/CD123 above the median in regenerating BM had a shorter RFS ( p < 0.001). The evolution of a ‘universal’ MRD marker is growing, with a similar approach being investigated by the EuroFlow Consortium to use single-tube disease-oriented panels for MRD detection over patient-specific antibody combinations.39 This concept of a standardised MRD marker panel will be pivotal in establishing consistency in MRD analysis. ‘Different from normal’ The ‘different from normal’ approach to MRD analysis overcomes the problem of MRD false negatives due to immunophenotypic shifts. Identification of MRD is reliant on comparing patterns of normal antigenic expression through all stages of maturation against the sample of interest. Populations are considered abnormal if they are 0.5 decades away from the position of the corresponding normal cells.8,40 Another advantage to this approach compared to LAIP identification is that MRD analysis is still possible despite not having an available diagnostic LAIP, which may occur when samples are sent to a centralised laboratory for evaluation. This method of analysis necessitates expert knowledge of normal haematopoietic maturation patterns, however, which may be a major limiting factor in its widespread implementation in laboratories. Achieving standardisation in MRD analysis when applying this approach is also more complex, as the determination of ‘normal’ patterns can be somewhat subjective. Each laboratory should establish normal expression with their antibody panel using a minimum of 20 patient/donor samples. We recommend both methods be routinely used when performing MRD analysis. However, this approach significantly increases the time spent per evaluation, therefore workflow and efficiency requirements of diagnostic laboratories should be considered before a laboratory offers MRD testing.
STANDARDISATION Definition of a LAIP One of the key difficulties in standardising MPFC-based MRD monitoring stems from a lack of consensus regarding the true definition of a LAIP. Different groups adopt conflicting criteria resulting in variably reported LAIPs. According to Feller et al.10 and Terwijn et al.41 a LAIP was considered valid only if expressed on at least 10% of leukaemic blasts. Alternatively, Buccisano et al.42 defined a LAIP to be expressed on at
Copyright © Royal College of pathologists of Australasia. Unauthorized reproduction of this article is prohibited.
FLOW CYTOMETRY IN ACUTE LEUKAEMIA Diagnostic AML
[Singlets]
103
[WBCs]
103
[CD45 progenitor gate]
103
CD117+
Viable cells
102
101
102 CD117 PC5
CD45 V500-A
102 SSC-A
611
101 CD45 progenitor gate
100
100
0
0
101 2 0
200
400 600 FSC-A [CD117+]
800
0
1000
10
1
10 SSC-A [CD11b-]
2
10
10
3
102
103
102
102
102
1 102 101 HLA-DR APC-H7
103
CD33 V450
103
5
101 SSC-A
[CD13+ dim]
103
101
100
0
103
CD13 PE
CD11b APC
0
–2 0
101
101 CD13+ dim
2
100
0 0
0.2 0
–5 –2 –4 –2
0
2
4
101 CD34 PE-Cy7
102
–4 –2
103
0
2
4 101 CD34 PE-Cy7
102
–1
103
0
MRD AML [WBCs]
[Singlets]
[CD45 progenitor gate]
103
103
103
CD117+
Viable cells 2
10
101
100
CD117 PC5
102 CD45 V500
SSC-A
102
101
101
CD45 progenitor gate
100
2
0 –0.5
0
0 0
0
200
400 600 FSC-A
800
1
0
10 SSC-A
10
1000 Gate All
2
10
10
3
–2
Number
[CD11b-] [CD117+] 10
101 5
101
101 100
100
MRD
0
–2
0
101 2 CD34 PE-Cy7
102
103
0
CD13+ DIM
0.2 0
–5
103
102 CD33 V450
CD13 PE
CD11b APC
CD11B-
102
103
102
102
101 SSC-A
[CD13+ dim] HLA-DR APC-H7 / CD33 V450
103
3
100
0
390,895
–1 –2
0
101 2 CD34 PE-Cy7
102
0
1
101 102 HLA-DR APC-47
103
103 Gate number All 1,152 MRD 1,152
Fig. 1 Diagnostic AML: Initial sequential gating strategy to include viable cells only (not shown): event number vs time, FSC-A vs FSC-H and FSC-A vs SSC-A. LAIP comprises abnormal antigen expression compared to normal blasts (CD34, dim CD13þ, partial HLA-DR). MRD AML: A sequential gating strategy is used in combination with physical characteristics (SSC-A vs FSC-A) to identify minimal residual disease (MRD). MRD is discriminated from promyelocytes (shown in purple) by their physical characteristics, high side scatter, HLA-DR positivity and bright CD13 expression, compared with dim CD13 expression on the leukaemic blasts. Normal blast expression (arrow) is confirmed by comparing to blasts from a healthy control. MRD ¼ 0.29% (MRD events/total WBCs 100). Black, leukaemic myeloblasts/MRD; purple, promyelocytes. WBC, white blood cells.
Copyright © Royal College of pathologists of Australasia. Unauthorized reproduction of this article is prohibited.
612
LEE et al.
least 50% of blasts. Different still, Olaru et al.43 deemed a LAIP to be significant if it was expressed on at least 1% of the total population (comprising 100,000 CD45 positive events). Denominator Discrepancy amongst authors also exists regarding the denominator used in MRD calculations. The most common denominator used is the total white blood cell (WBC) count. The lysing reagent eliminates the majority of red cells during sample preparation, (although this process can be imperfect) before further exclusion by a CD45 vs side scatter gating strategy. However, in paediatric ALL the denominator is mononuclear cells, which is based on the tradition of using ficoll to extract the mononuclear component. Although both techniques were recognised as sample preparation methods at the Proceedings of the Second International Symposium on MRD assessment,44 the recently published British Committee for Standards in Haematology (BCSH) Guidelines recommend red cell lysis over density centrifugation.45 Number of events to acquire The total number of events recommended for acquisition ranges from 100,00043 to 250,0007,46 to 1 106.41,44 Variability amongst groups regarding what constitutes an ‘event’ also influences sensitivity with some groups aiming for total events, which would include debris and erythroid precursors, and others aiming for total WBC events. Both of these methods are open to errors in enumeration due to peripheral blood contamination in bone marrow aspirate samples.47 Attaining adequate event detection can also be problematic in paucicellular marrows. Techniques to address these limitations include doubling the standard volume of cells, density gradient centrifugation or ‘bulk lysis’.48 Sample quality can also be assessed by ensuring adequate levels of cells that are normally restricted to the marrow such as erythroid, myeloid and B-cell progenitors are present, a technique recommended by the European Myeloma Network when performing myeloma MRD.49 The number of events that defines a leukaemic population also varies significantly between diseases and groups. The IBFM-ALL-FLOW MRD Network recommends a minimum of 30 leukaemic events per tube accompanied by a confirmatory, independent tube.44 The EuroFlow Consortium recommends at least 100 leukaemic cells (as the sum of events in all tubes) in line with their recommendation for MPFC MRD quantitation of multiple myeloma.49 In our laboratory we aim to acquire >500,000–1,000,000 events in AML and ALL. The appropriate absolute cut-off of abnormal leukaemic cells varies on an individual patient basis, but 50–100 abnormal events would usually ensure a sensitivity around 1 in 10,000. Reproducibility The perceived lack of reproducibility of MPFC, particularly across multicentre studies, has so far hindered the progress of standardisation.39,50 Reproducibility is influenced by a large number of complex factors including equipment and reagent variability and stability, software, and data analysis and interpretation. At the heart is a thorough and robust quality framework. The International Council for Standardisation of Haematology (ICSH), the International Clinical Cytometry Society (ICCS) and the BCSH have released independent
Pathology (2015), 47(7), December
practice guidelines in an attempt to address the issues of standardisation and promote inter-laboratory harmonisation.45,51 The recommendations provide guidance on instrument set up, fluorochrome and panel design, sample preparation, data analysis and quality control. It is hoped these guidelines will improve uniformity of practice across laboratories for all diagnostic cases. Notably, MRD analysis was not addressed in any of the aforementioned guidelines. The EuroFlow Consortium published recommendations for MRD markers in antibody panels, but also did not specifically address standardisation of MRD analysis. Feller et al.52 demonstrated that standardisation of LAIPs for MRD analysis is possible in a multicentre setting. This was achieved by defining a standardised antibody panel and standard operating procedures. Newer developments such as automated analysis algorithms may also be helpful in improving uniformity.53
ACUTE MYELOID LEUKAEMIA Prognostic relevance of MRD following induction and consolidation San Miguel et al.9 were the first to demonstrate the prognostic significance of flow-based MRD in AML, identifying two levels of MRD (0.5% and 0.2%, respectively) that correlated with a higher risk of relapse post-induction and post-intensification. In a later study by the same group, the authors showed that post-induction MRD levels could also stratify patients into different risk groups.11 Buccisano et al.42 demonstrated that post-consolidation MRD levels >0.035% correlated with an unfavourable prognosis in terms of relapse rate, OS and RFS ( p < 0.001). In contrast to San Miguel et al.,11 they found that post-induction MRD status bore no influence on outcome, provided MRD negativity was achieved by the end of consolidation.42 Postconsolidation MRD levels proved to be an independent predictor of outcome when compared against karyotype, age and post-induction MRD levels as covariates. Effect of chemotherapy on MRD The type of chemotherapy administered may affect the level of MRD achievable and re-calculation may be necessary for treatment protocols that involve more intensive regimens resulting in greater tumour debulking and potentially less MRD.4,9,42 The findings of Rubnitz et al.54 and Feller et al.10 would contradict this suggestion where treatment regimens did not appear to have any impact on the degree of MRD positivity. Timing The optimal time-point for MRD assessment is still unresolved with some advocating post-induction,8,11,41 post-consolidation,6,42 or both.7,10 Timing is critical when applying riskadapted therapy. Premature evaluation would potentially identify more patients resulting in additional (and unnecessary) intensive therapy. However, deferred evaluation (i.e., postconsolidation) may have limited practical utility. While there is no consensus regarding the optimal time-point for MRD assessment, earlier evaluation (post-induction) is likely to be the most useful so that treatment could be altered if required.1,5,8,11,41,54,55
Copyright © Royal College of pathologists of Australasia. Unauthorized reproduction of this article is prohibited.
FLOW CYTOMETRY IN ACUTE LEUKAEMIA
Also unclear is the frequency of MRD assessments. Although regular monitoring may provide peace of mind, this would involve frequent and possibly unnecessary bone marrow biopsies, which are not without risks. Intimately associated with this question is whether earlier therapy (i.e., in the absence of morphological relapse) ultimately improves survival. We look forward to prospective, clinical trials that may adequately answer these questions. Quality of response The kinetics of blast reduction has been found to be independently associated with OS. The German AML Cooperative Group determined that morphologically detectable blasts midinduction indicated slow clearance and was a negative predictor of event-free survival (EFS), OS and RFS.56 This has been reaffirmed in later trials demonstrating early blast clearance as an independent predictor of survival post-induction.57,58 Maurillo et al.59 disagreed with these findings, suggesting that achievement of MRD negativity by the end of consolidation is more important than achieving early MRD negativity. They demonstrated that 30% of patients who were MRD positive post induction went on to achieve MRD negativity following consolidation, with no significant difference in RFS and OS between MRD negative patients post-induction and consolidation and MRD positive patients post-induction who converted to MRD negative status following consolidation. This would be concordant with the findings of Buccisano et al.42 and Venditti et al.,6 suggesting the kinetics of blast clearance may not be an accurate predictor of survival, which seems counter-intuitive. Alternatively, Rubnitz et al.54 suggested that the depth of response might be more predictive of outcome. Here, a riskadaptive approach incorporating serial MRD assessments (following induction 1 and 2) was applied to guide therapy intensification in a paediatric AML cohort.54 The outcome for patients with low levels of MRD (0.1% to 0.1% post-induction 1 and 2 was an independent predictor of OS, compared with cytogenetics and white cell
count (WCC) ( p ¼ 0.037, HR 3.79, and p ¼ 0.022, HR 6.15, respectively). These findings and their approach to MRD assessment subsequently informed the risk assignment algorithm in the multicentre AML2002 study, which was discussed earlier.54 Sievers et al.8 demonstrated that patients with MRD >0.5% after remission induction were 4.8 times more likely to relapse ( p < 0.0001) with a mortality rate that was 3.1 times greater ( p < 0.0001). Following consolidation, OS at 3 years was 41% vs 69% for patients with >0.5% vs 30 109/L, t(4;11) or other MLL rearrangement) were assessed for MRD by 4-colour flow cytometry to guide treatment stratification. Patients with MRD levels
