Acute promyelocytic leukemia (APL) is a subtype of acute myeloid leukemia (AML) with specific biology and clinical presentation. The vast majority of cases are characterized by a translocation that fuses the promyelocytic leukemia gene (PML) on chromosome 15 with the gene-encoding retinoic acid receptor-alpha (RARα) on chromosome 17 (1). Because APL can be rapidly fatal, with 5–10% of newly diagnosed patients succumbing to hemorrhagic death, and because most patients can be cured with early administration of therapy, sensitive and specific methods to screen AML cases before the biomolecular confirmation of PML/RARα rearrangement are needed (2).
Morphologic evidence of hypergranular leukemic promyelocytes generally implies the presence of APL cells and justifies immediate treatment initiation (2, 3). However, more robust diagnostic procedures are represented by cytogenetic, fluorescence in situ hybridization (FISH), and reverse transcriptase–polymerase chain reaction (RT-PCR) in order to precisely identify the PML-RARα fusion gene, but they can be time consuming and require specialized laboratories (3, 4). Alternative diagnostic procedures for APL diagnosis include PML pattern recognition by immunofluorescence microscopy (4) and surface immunophenotypic analysis by conventional flow cytometry (FCM) (5).
Immunofluorescence analysis is a rapid and specific diagnostic tool to distinguish APL blasts that exhibit multiple diffuse small PML bodies from cells with wild-type distribution consisting of no more than 20 distinct large PML bodies (4). However, this technique can involve operator biases and has limited statistical power. Although conventional FCM has shown documented efficacy for a rapid diagnosis of APL by the identification of APL blasts with pathological immunophenotype, confirmation of the PML pattern associated with APL is not possible by this technique (5, 6).
Recently, Grimwade et al. (7) were able to quantify PML protein pattern expression at diagnosis using the ImageStream imaging FCM on a cohort of 18 AML patients, 4 of which were APL with a blast range between 75 and 85%. In wake of their work, we extended the strategy of ImageStream APL identification by showing rare event detection in dilution experiments. Our work was based on the fact that imaging FCM combines high-speed image capture with image quantification, enabling statistical and objective discrimination of cells based on their appearance (8). In this way, we were able to: i) identify single live cells in fixed/permeabilized cellular suspensions; ii) determine “micro-” and “macrospeckled” PML expression pattern in NB4 and HL60 cell lines as well as in fresh APL and normal bone marrow (BM) cells, respectively; iii) demonstrate the ImageStream ability to detect “microspeckled” pattern up to 10% APL cell presence in electronic and real dilution experiments spiking NB4 in the midst of HL60 as well as fresh APL blasts in normal BM cells.
Materials and Methods
HL-60 (human AML FAB-M2) and NB4 (human APL FAB-M3, PML/RARα+) cell lines were acquired from Deutsche Sammlung von Mikroorganismen und Zellkulturen and maintained in continuous culture at CEINGE-Biotecnologie Avanzate Cell Culture Facility (Naples, Italy) according to UKCCCR guidelines for the use of cell lines in cancer research (9). Particularly, both cell lines were cultured in 24 well plates at 500,000 cells/mL in RPMI (SIGMA-ALDRICH, St. Louis, MO) supplemented with 10% heat-inactivated fetal bovine serum (LONZA Basel, Switzerland) and 1% Ultraglutamine (LONZA Basel, Switzerland).
Normal BM sample was obtained from a patient with non-Hodgkin lymphoma who underwent BM aspiration in the context of routine clinical practice. APL BM was obtained from a patient with 82% of leukemic promyelocytes at diagnosis. In both cases, BM sample was obtained following informed consent.
Successively, APL and normal BM cells were obtained after red blood cell lysis with NH4CL. After lysis, cells were counted with Invitrogen Countess™ automatic cell counter (Life Technologies, MI, Italy), and cellular viability was assessed by Trypan blue exclusion. Finally, cellular dilutions were performed as described in Table 2, and a total of 500,000 cells were used for each ImageStream test.
Cellular Fixation, Permeabilization, and Staining
Cells were fixed and stained with BD Cytofix/Cytoperm kit (BD Biosciences, San Jose, CA) according to the manufacturer's instructions. Fixed/permeabilized cells were diluted in 100 μL of BD Perm/Wash solution containing 30 μL of AlexaFluor488-conjugated anti-PML (clone PG-M3; Santa Cruz Biotechnology, Santa Cruz, CA) monoclonal antibody and incubated for 90 min at room temperature. Unstained fixed/permeabilized cells were used as negative staining control. After incubation, cells were washed twice with 1 mL 1× BD perm/wash solution and stained with 50 nM of DRAQ5 (Biostatus, UK), a cell permeable far-red fluorescent DNA dye.
All samples were run on the ImageStream100 flow cytometer (Amnis Corp., Seattle), and 20,000 cells per sample were collected with 150 mW 488-nm laser power with brightfield set to channel 5. Only events with a minimum bright field area of 12 μm2 or greater were included in the data file to eliminate collection of small debris. Data were acquired using the INSPIRE (Amnis Corp.) software and analyzed by the IDEAS software (Amnis Corp.).
Single, nonapoptotic cells in BD Cytofix/Cytoperm fixed cellular suspensions were identified as shown in Figure 1. First, single cells included in the blue region (Fig. 1, panel A) with normal DNA content and high-aspect (width/height) ratio were discriminated from multicellular events (high-DRAQ5 intensity and low-aspect ratios) and debris (low-DNA content). Nonapoptotic cells were then selected in the yellow region on the DRAQ5 threshold area versus SSC mean pixel (Fig. 1, panel B). Condensed and fragmented nuclei of apoptotic cells displayed lower nuclear areas compared to live cells; apoptotic cells often had higher scatter as well. Finally, within the live gate, the cells of best focus (cell in focus gate, not shown) were gated for further analysis.
Qualitatively, two distinct “micro-” and “macrospeckled” PML expression patterns were seen in NB-4 and HL-60 cell lines as well as in fresh APL and normal BM cells, respectively. To quantitatively distinguish these patterns, two features were used in the analysis: Spot Count and Modulation. The Spot Count feature enumerates bright PML spots per cell. Using this parameter, we found that HL-60 cells, which displayed a predominantly “macrospeckled” phenotype, had much higher mean spots per cell compared to NB4 cells, which displayed a predominantly “microspeckeld” pattern (3.52 vs. 0.42 mean spots/cell, Fig. 2, Panel A and B, respectively). We also used the Modulation texture parameter to measure PML distribution. Modulation = (max pixel – min pixel)/(max pixel + min pixel). The bigger the range in the per-cell PML stains, the higher the modulation scores. Because of the high range in pixel intensity associated with “macrospeckled” PML distribution, HL60 cells had significantly higher modulation values than NB4 cells displaying a “microspeckled” pattern (Fig. 3). Additionally, we evidenced the “microspeckled” pattern also in case of fresh APL blasts using the modulation feature and the “macrospeckled” one in case of normal BM cells (Fig. 4).
The second step of our work was to see if it was possible to detect limiting numbers of APL cells within a larger cell population, we performed preliminary electronic mixing experiments of two cell lines such as NB-4 and HL-60. Briefly, randomly selected events from the NB4 file were combined with the HL60 file to form merged files that contained a specific ratio of HL60 and NB4 events. Triplicate files with 999/1 and 99/1 HL60/NB4 ratios, along with a single file with 9/1 and 1/1 HL60/NB4 ratios, were used (Table 1). These merged files were then analyzed with the same template to obtain PML modulation mean scores (Table 1). The relative percentage of NB4 cells detected within the merged files was estimated by calculating the increase in the mean feature values over negative control (F0 = 100% HL60 sample) and normalizing to the mean feature value range between negative and positive controls (F100 = 100% NB4 sample). This normalized percentage (Fn) was computed as follows: Fn = 100 × (Fm – F0)/(F100 – F0), where Fm is the measured feature value. In this way, the percentage of NB4 contamination was determined by modulation normalization as follows: Fn (NB-4) = 100 × (Fm – 0.4479)/(0.2372 – 0.4479). Both the non-normalized and normalized modulation scores were reported in Table I. Importantly, by the use of normalized modulation feature, we were able to identify up to 10% of virtual contamination for NB-4 cells (Table 1).
Table 1. The non-normalized and normalized PML mean modulation normalization scores in electronic dilution experiments
PML modulation, mean
% NB4, modulation normalized
ImageStream cytometer is able to identify NB-4 cells in 10% virtual dilution experiments by normalized modulation, as evidenced in bold.
By virtue of these in silico tests, we evaluated the ability of ImageStream FCM to detect “microspeckled” pattern in real dilution experiments mixing NB4 in HL60 cells as well as fresh APL blasts in normal BM cells at 50%, 10%, 1%, and 0.1% levels. As evidenced in Table 2, APL cellular contaminations were detected up to 10% level using the mean modulation feature in both NB4 mixed in HL60 as well as fresh APL mixed in normal BM cells. Dilution at 1% and 0.1% levels was nondetected, and the mean modulation was closely similar to that evidenced for uncontaminated HL60 cell line and normal BM cells (data not shown). Finally, using the mean modulation normalization feature, we were able to find up to 10% of APL-contaminating cells with respect to the performed dilutions, as evidenced in Table 2.
Table 2. Identification of “Microspeckled” pattern in real dilution experiments
PML modulation, mean
Mean modulation normalized (%)
PML mean modulation is able to evidence PML pattern alteration in NB4 and fresh APL blasts. Real dilution experiments evidenced APL leukemic presence up to 10% dilution. Mean modulation normalization values were useful to evaluate the percentage of contaminating APL cells.
APL is a medical emergency frequently presenting with an abrupt onset (1). The high risk of early hemorrhagic mortality, which still accounts for 5–10% of newly diagnosed patients, and the potential for high-cure rate (>80%) highlight the importance of immediate recognition and prompt initiation of specific treatment (2). The identification of the APL-specific genetic lesion in leukemic cells is feasible at chromosome, DNA, RNA, and protein levels with the use of conventional karyotyping, FISH, RT-PCR, and anti-PML monoclonal antibodies, respectively. Compared to conventional karyotyping, RT-PCR and FISH do not require dividing cells for analysis, and they allow results to be obtained in cases where the PML-RARα fusion gene is formed as a result of cryptic or complex rearrangements in the absence of the classic t(15;17) (3). However, because both these techniques are technically challenging and RT-PCR is notoriously prone to contamination and artifacts (3), it is advisable that diagnostic and follow-up samples are sent to reference laboratories where well-trained personnel have specific experience with PML/RARα (3). Therefore, a more rapid initial diagnosis procedure is required.
Rapid diagnostic techniques use microscopic and/or flow cytometric analysis of patient blood and BM samples. The morphologic appearance of hypergranular leukemic promyelocytes allows typical cases to be identified and justifies immediate treatment initiation, without waiting for diagnostic confirmation at the genetic level (3). In addition, cells with microgranular PML nuclear distribution associated with PML/RARα can be readily distinguished from other leukemic and normal hematopoietic cells having an aggregated PML nuclear pattern by using immunofluorescence microscopy (4). In light of its very convenient cost-benefit ratio, this assay is highly recommended to rapidly confirm diagnosis of APL at the protein level, especially in small institutions not equipped and experienced for genetic analyses (10). However, traditional imaging techniques suffer from poor statistics and lack of standard quantitative metrics, making definitive identification of APL difficult in some cases.
Conventional nonimaging FCM also plays an important role for the diagnosis and monitoring of APL cells (5, 11). Indeed, APL is characterized by a highly specific immunophenotype showing a consistent absence or very low expression of CD15, CD11a, CD11b, CD11c, CD18, CD66b, and CD66c molecules (5, 12). At the same time, the blasts exhibit a wide range of CD13 expression (broad histogram), whereas CD33 expression is homogeneous (sharp histogram). CD34 and human leukocyte antigen (HLA)-DR are frequently absent (5). This immunophenotype, in skillful hands, has been reported to show high sensitivity and specificity for predicting APL molecular rearrangement (12, 13).
The emergence of high-speed imaging FCM has allowed exploration of new frontiers in clinical hematology, particularly in case of APL diagnosis, as demonstrated in this work. Indeed, ImageStream technology combines a precise method of electronically tracking moving cells with a high-resolution multispectral imaging system to acquire multiple images of each cell in different imaging modes (8). The current commercial embodiment simultaneously acquires up to 12 high-resolution images of each cell at high rates of capture, with fluorescence sensitivity comparable to conventional FCM (8). Thus, ImageStream cytometry can combine the “power of surface hematology” (5) with the quantitative analysis of PML protein distribution.
In this work, we evaluated the feasibility of ImageStream as a new potential diagnostic tool useful for a rapid, accurate, and operator-independent APL diagnosis. Indeed, we were able to differentiate the HL60 “macrospeckled” and NB4 “microspeckled” PML pattern by ImageStream FCM using Modulation parameter. Our finding extends the recent work of Grimwade et al. (7), by demonstrating the feasibility of using ImageStream cytometry for the detection of rare APL events in electronic and real dilution experiments. For the first time, by quantifying PML distribution using modulation parameter, we were able to clearly detect PML pathological pattern even at 10% APL cell presence in both NB-4 mixed in HL-60 cell lines as well as fresh APL blasts in normal BM cells. Furthermore, because, as reported by Dimov et al. (14), the APL blast percentage at diagnosis ranges from 21 to 97%, ImageStream FCM could be particularly helpful for APL cases where pathological promyelocytes are difficult to detect, and the laboratory hematologist interest is to confirm a suspected diagnosis.
Although future studies are needed and molecular genotypic analysis has the final say, our data importantly indicate that imaging FCM may be a frontline test for rapid confirmation of suspected APL increasing the laboratory hematologist ability to rapidly identify even rare APL cells.
Ideally, combining surface immunophenotyping with quantitative analysis of PML distribution pattern analysis in a single platform analysis would be attractive for hematologists desiring an operator-independent, prompt, and robust APL diagnosis with 100% sensitivity and specificity.
The authors gratefully acknowledge the constant support of the Biobank staff of CEINGE-Biotecnologie Avanzate di Napoli. The authors Peppino Mirabelli, Giulia Scalia, Caterina Pascariello, Francesca D'Alessio, Elisabetta Mariotti, and Rosa Di Noto give special thanks to Professor Francesco Salvatore for his constant guidance and help in professional growth.
Note Added in Proof
PM, GS and RK prepared the cells, designed the experiments and performed flow cytometry assays.
CP, FDA and EM participated in study design, data analysis and revision of the manuscript.
TG and VV participated in data analysis and manuscript preparation.
DB and LDV conceived the study, designed the experiments and wrote the manuscript together with RDN.