Brief Report
Detection of Cytoplasmic Nucleophosmin Expression by Imaging Flow Cytometry Lizz Grimwade,1* Emma Gudgin,2 David Bloxham,1 Graham Bottley,3 George Vassiliou,4 Brian Huntly,2 Mike A. Scott,1 Wendy N. Erber1,5
1
Haemato-Oncology Diagnostics Service, Department of Haematology, Addenbrooke’s Hospital, Cambridge, United Kingdom
2
Department of Haematology, Cambridge Institute for Medical Research, University of Cambridge, Cambridge, United Kingdom
3
Cronus Technologies, Camberley, United Kingdom
4
Haematological Cancer Genetics, Wellcome Trust Sanger Institute, Cambridge, United Kingdom
5
Pathology and Laboratory Medicine, University of Western Australia, Australia
Received 9 February 2012; Revision Received 22 May 2012; Accepted 24 July 2012 Additional Supporting Information may be found in the online version of this article. *Correspondence to: Lizz F. Grimwade, Department of Haematology, Box 234, Addenbrooke’s Hospital, Hills Road, Cambridge CB2 0QQ, United Kingdom Email:
[email protected] Published online 11 September 2012 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/cyto.a.22116 © 2012 International Society for Advancement of Cytometry
Cytometry Part A 81A: 896 900, 2012
Abstract Mutations within the nucleophosmin NPM1 gene occur in approximately one-third of cases of acute myeloid leukemia (AML). These mutations result in cytoplasmic accumulation of the mutant NPM protein. NPM1 mutations are currently detected by molecular methods. Using samples from 37 AML patients, we investigated whether imaging flow cytometry could be a viable alternative to this current technique. Bone marrow/peripheral blood cells were stained with anti-NPM antibody and DRAQ5 nuclear stain, and data were acquired on an ImageStream imaging flow cytometer (Amnis Corp., Seattle, USA). Using the similarity feature for data analysis, we demonstrated that this technique could successfully identify cases of AML with a NPM1 mutation based on cytoplasmic NPM protein staining (at similarity threshold of 1.1 sensitivity 88% and specificity 90%). Combining data of mean fluorescence intensity and % dissimilar staining in a 0–2 scoring system further improved the sensitivity (100%). Imaging flow cytometry has the potential to be included as part of a standard flow cytometry antibody panel to identify potential NPM1 mutations as part of diagnosis and minimal residual disease monitoring. Imaging flow cytometry is an exciting technology that has many possible applications in the diagnosis of hematological malignancies, including the potential to integrate modalities. ' 2012 International Society for Advancement of Cytometry Key terms imaging flow cytometry; acute leukemia; nucleophosmin; hematology
NUCLEOPHOSMIN (NPM)
is both a nuclear (predominantly nucleolar) and cytoplasmic protein that functions as a molecular chaperone (1). It has many functions including the prevention of protein aggregation in the nucleolus and regulation of p53 levels (2). Mutations within the C-terminal region (exon 12) of the NPM1 gene occur in approximately one-third of cases of de novo acute myeloid leukemia (AML) (3,4) leading to increased nuclear export and aberrant cytoplasmic accumulation of the mutant NPM protein (5). A number of methods have been assessed to detect cytoplasmic NPM (cNPM) in clinical specimens. Immunohistochemistry has been reported to be reliable, especially on B5-fixed bone marrow trephines, with close correlation between cNPM and the presence of mutated NPM1 (6). Immunocytochemistry of air-dried blood and bone marrow smears can also detect cNPM in myeloblasts but is insufficiently reliable for diagnostic use (7). Fluorescent confocal microscopy and standard flow cytometry have both been evaluated for cNPM detection (8,9). Standard flow cytometry cannot be used to assess the localization of molecules within specific cellular compartments, but two alternative approaches have been assessed. The first, using antibodies to mutant NPM1, was able to identify NPM1-mutated AML cases exclusively with or without the addition of a de novo NPM ‘‘control’’ antibody (8,9). The second method specifically addressed mean fluorescence intensity (MFI) and showed higher MFI in NPM1 mutated than wild-type AML cases (10). However, discrepancies between flow results and NPM1 mutation status were seen in both these studies.
BRIEF REPORT Table 1. Determination of similarity threshold cut-off (selected results) THRESHOLD CUT-OFF
1.0 1.1 1.2 1.0 1.1 1.2
% OF DISSIMILAR CELLS AS ‘‘POSITIVE’’
PPV
NPV
45 45 45 50 50 50
0.875 0.882 0.882 0.875 0.882 0.895
0.857 0.857 0.9 0.857 0.9 0.83
SENSITIVITY SPECIFICITY (%) (%)
83 83 88 83 88 88
90 90 90 90 90 85
Various similarity threshold cut-offs and similarity percentages for mutated NPM staining were assessed (55 and 60% dissimilar and 0.9 and 0.95 threshold cut-offs were also assessed, data not shown) to determine the best combination of positive predictive value (PPV), negative predictive value (NPV), sensitivity, and similarity.
Recently we reported success using imaging flow cytometry to assess the intracellular localization and pattern of disrupted PML bodies in AML (11), which was subsequently confirmed by another research group to be a reliable technique (12). We therefore explored the applicability of this technology for discriminating cytoplasmic from nuclear NPM in AML.
MATERIALS AND METHODS Forty-nine newly diagnosed patients with AML who presented in 2010–2011 were selected for analysis and 37 were included in this project (Table 2). Eight samples were excluded pre-analysis because of poor quality or low-cell numbers. All patients had given written consent for their samples to be assessed, and the study was approved by the appropriate ethics committee. Seventeen of the 37 cases had an NPM1 mutation (performed by high-resolution melt analysis) (13). Fresh peripheral blood or bone marrow cells were stained using an anti-NPM antibody [clone NA24, kindly provided by the University of Oxford via personal communication (1)] followed by FITC-conjugated goat anti-mouse-immunoglobulin (catalog F0232, Dako, UK). Staining was performed using an intracellular fixation and permeabilization method with CD45PE (clone Immu-19.2, catalog IM1833, Beckman Coulter, UK) and DRAQ5 (catalog DR71000, Biostatus, UK) nuclear stain as described previously (11). Data from a minimum of 5,000 cells were acquired using the 488-nm solid-state laser (40 mW) on an ImageStream IS100 imaging flow cytometer (Amnis, USA). Compensation settings for all analyses were set on single-color controls for all antibodies/fluorochromes. Acquired data were analyzed using the IDEAS analysis software. Single-focused blast cells were gated as described previously using side-scatter and CD45 expression, and dualpositive cells were selected accordingly by examination of digital images and incorporating values from the single-color control samples (11; see Supporting Information). Cells were also visually examined using the brightfield images to check for significant cell degeneration, which may have occurred during the staining process, which would affect results; no samples Cytometry Part A 81A: 896 900, 2012
were excluded from the study based on cell degeneration. The NPM-staining pattern was assessed using the similarity analysis feature within the software using the mask Dilate [Morphology (M06, DRAQ-5; 1)]. ‘‘Masks’’ are a key component of the IDEAS analysis software and define a set of digital image pixels that contain the region of interest within a cell. Masks can be used alone or combined. The Dilate [Morphology (M06, DRAQ-5; 1)] mask that refers to the nuclear DRAQ5 signal was used for the similarity analysis. The morphology mask was used to determine the nuclear region of interest within the cell, and this was then fine-tuned by adding a ‘‘dilate’’ mask to dilate the morphology mask by one pixel to further define the nuclear area to ensure that some cytoplasmic pixels were included in the analysis (see Supporting Information). From a primary training set of 10 patients, a threshold of 1.1 was found to give the best discrimination between positive and negative patients and therefore chosen for subsequent analysis. At this 1.1 threshold cut-off, if 51% of cells showed ‘‘dissimilar’’ staining, that is, different localization of FITC and DRAQ5, this was considered a ‘‘positive’’ result, representing cNPM localization. The digital images were used to visually determine and confirm the appropriate cut-off values and distinguish between positive and negative events. The threshold and ‘‘positive’’ percentage values were determined by looking at different cut-off values and ‘‘positivity’’ percentages to give the best discrimination between positive and negative patients based on assessing positive and negative predictive values, specificity, and sensitivity (Table 1). An analysis template was used to ensure that all variables/settings were the same for all samples.
RESULTS There were significant differences in the percentage of cells with ‘‘dissimilar’’ staining at a similarity threshold of 1.1 between AML patients with wild-type NPM1 and those with the NPM1 mutation (P \ 0.0001; Table 2 and Figs. 1B–1D). These differences were clearly visible in the captured digital images. ‘‘Similar’’ staining showed colocalized NPM-FITC signal and nuclear DRAQ5 (Figs. 1E and 1F). ‘‘Dissimilar’’ patterns had a clear distinct cNPM-positive fluorescent ring around the DRAQ5-positive nucleus (Figs. 1G and 1H). Both nuclear and cNPM stain were visible in NPM11 AML cases (Figs. 1H and 1I). Overall, 15 of 17 AML NPM1-mutated cases had greater than 51% ‘‘dissimilar’’ cell staining at the 1.1 similarity threshold and were classed as ‘‘positive’’ for cNPM localization (sensitivity 88%, specificity 90%; Fig. 1B). Discrepant results were seen: two cases with NPM1 mutations had ‘‘dissimilar’’ levels of \50% (Figs. 1J and 1K) on similarity analysis (i.e., ‘‘false-negative’’ results). However, the digital images of these two cases showed the vast majority of cells to have clear cytoplasmic as well as nuclear NPM stain. Furthermore, 2 of 20 NPM1-wild-type cases had ‘‘dissimilar’’ levels 51% (Figs. 1L and 1M) on similarity analysis (i.e., ‘‘false-positive’’ results); digital images confirmed the presence of cNPM in the majority of blast cells. The MFI of the total NPM-FITC signal in the AML NPM1-mutated cases was significantly 897
BRIEF REPORT Table 2. Summary of patient characteristics at diagnosis GENDER
AGE
SAMPLE TYPE
BLAST COUNT
NPM1 STATUS
CYTOGENETICS
MOLECULAR
Male Female Mean (years) Range (years) Bone marrow Peripheral blood Mean (%) Range (%) Wild type NPM1 mutated Normal Complex* Trisomy 8 BCR-ABL t(6;9);DEK-NUP214 Trisomy 5 t(8;21);AML1-ETO 12(mar) Not available FLT3 ITD FLT3 TKD
20 17 60.97 20–94 23 14 65.1 20–95 20 17 20 3 1 1 2 1 1 1 7 11 (1 biallelic) 3
IMMUNOPHENOTYPING RESULTS
ANTIGEN
ALL PATIENTS (N 5 37)
NPM1 MUTATED PATIENTS (N 5 37)
CD7 CD13 CD14 CD19 CD33 CD34 CD41 CD56 CD61 CD64 CD79a CD117 HLA-DR MPO
8 (22%) 35 (95%) 4 (11%) 1 (3%) 37 (100%) 26 (70%) 1 (3%) 3 (8%) 1 (3%) 14 (38%) 1 (3%) 35 (95%) 34 (92%) 18 (49%)
5 (14%) 15 (88%) 3 (18%) 0 (0%) 17 (100%) 9 (53%) 0 (0%) 1 (6%) 0 (0%) 9 (53%) 0 (0%) 16 (94%) 13 (35%) 15 (88%)
NPM1 STATUS
MFI MEAN (RANGE)
% DISSIMILAR STAINING MEAN (RANGE)
Wild-type Mutated
4,155 (1,091–12,003) 7,353 (1,544–12,558)
22.5% (2–82%) 60.2% (23–83%)
ITD, internal tandem duplication; TKD, tyrosine kinase domain; complex, 3 abnormalities on karyotyping; HLA, human leucocyte antigen; MPO, myeloperoxidase; MFI, mean fluorescence intensity. Immunophenotyping results are expressed as number of patients with positive results (% of total patients). No significant differences between the wild-type and NPM1 mutated patients groups were seen for gender, age, sample type, or blast count.
higher than the AML NPM1 wild-type cases (P 5 0.009; Table 2 and Fig. 1A). A scoring algorithm was then devised based on the actual % dissimilar staining (\20%, 20–50%, and 51%) and MFI (\4,500, 4,500; Table 3). Scores of 0, 1, or 2 were allocated that would give totals of 2 for ‘‘positive’’ and \2 for ‘‘negative’’ cases. When the scoring system was applied to NPM1mutated AML, all cases had a score of 2. This included the ‘‘false-negative’’ cases from the previous analyses that both scored 2. The scores for the NPM1 wild-type AML cases were all either 0 or 1. However, the algorithm could not identify the ‘‘false-positive’’ cases of wild-type AML, which scored 3. Utilization of this scoring algorithm improved the sensitivity of the results from 88 to 100%. It is also of interest to note that three cases (2 NPM1 mutated, 1 wild-type) had dissimilar values very close to the percent cut-off (51%, 51%, and 49%, respectively); these were correctly ‘‘scored’’ using the algorithm and had MFI values that were clearly within the range for their mutation status (8,173; 12,314; and 2,557, respectively).
DISCUSSION The results of the present study show that imaging flow cytometry can successfully discriminate cytoplasmic from nuclear NPM and identify cNPM staining in cases of AML with the NPM1 exon 12 mutation. It demonstrates the additional power that can be achieved by being able to visually review the digital images of all events and thereby confirm the flow-cyto898
metric electronic data. The digital images confirmed that in 15 of 17 AML NPM1-mutated cases, the algorithms accurately identified 51% cells with cNPM, and these patients were classed as ‘‘positive’’ for cNPM localization based on dissimilarity alone. In the remaining two NPM1-mutated cases with low-‘‘dissimilar’’ levels, the blast cells had predominantly nuclear NPM staining, but there was also clear cytoplasmic staining in the majority (Figs. 1J and 1K); a significant proportion did, however, lack cNPM. This demonstrates that there can be variability in cytoplasmic localization of NPM between cells within an individual case, contrary to reports from bone marrow trephine biopsies that claim all cells to have cNPM (6). It is of note that the NPM MFI for these two cases was more in keeping with the NPM1-mutated AML average than wild-type AML (Fig. 1A). Using the scoring algorithm we created, incorporating both MFI and percent dissimilar staining, these two cases would be flagged as potential positives that would allow clinicians to consider therapeutic options whilst awaiting confirmation by molecular methods. The only discrepant results were 2 of 20 AML NPM1 wild-type cases that showed cNPM on imaging, statistical evidence of ‘‘dissimilar’’ (70 and 77%) staining, and scored 3 using the MFI/% dissimilar staining scoring algorithm. Manually, altering the similarity threshold from 0.9 to 1.2 did not bring these cases into the range of wild-type AML. We also investigated different ‘‘masking’’ strategies in combination with the similarity analysis feature, including masks that Detection of cNPM Expression by Imaging Flow Cytometry
BRIEF REPORT
Figure 1. Nucleophosmin analysis on the ImageStream. A: Comparison between NPM1 mutated and wild-type patients showing the variation in MFI of the NPM-FITC signal (purple, false-positive cases; green, false-negative cases on similarity analysis) and (B) percentage of cells with ‘‘dissimilar’’ staining at the 50% cut-off at a 1.1 similarity threshold (error bars signify standard deviation). Similarity analysis showing the cut-off at 1.1 (C) in a NPM1 wild-type AML patient and (D) in a NPM1 exon 12 AML patient. Fluorescein isothiocyanate (FITC)NPM (green) and DRAQ5 (pink) digital images showing wild-type AML similar staining patterns (E and F) and mutated NPM1 AML ‘‘dissimilar’’ staining pattern (G). In mutated cases (H), nuclear NPM staining is clearly visible in addition to cytoplasmic staining when the DRAQ5 signal is removed from the digital image (I). J, K: Two NPM1 mutated cases with \50% ‘‘dissimilar’’ all show clear cytoplasmic staining upon visualization. L, M: The two wild-type cases with 51% ‘‘dissimilar’’ were confirmed to have cytoplasmic staining on image analysis. Percentage values given with each image relate to the percentage of cells with dissimilar staining for the respective case.
incorporated more cytoplasmic pixels. These did not significantly affect the data and had no effect on the outlier patients. In spite of these outlier results, the negative predictive value of imaging flow cytometry method in excluding an NPM1 exon 12 mutation is 90%. We have no methodological or analytical explanation for these results (Figs. 1L and 1M). However, one of these patients had t(6;9);DEK-CAN, which involves nucleoporin, a protein responsible for maintaining transport of proCytometry Part A 81A: 896 900, 2012
teins through the nuclear membrane. Nucleoporin could potentially play a role in translocation of NPM to the cytoplasm. The second patient did not have a detectable genetic defect, and the explanation for the ‘‘false-positive’’ result is not known. The report by Oelschlaegel et al. (10) had also shown ‘‘false-positive’’ NPM results by flow cytometry (based on high MFI) in 6 of 298 NPM1 wild-type patients. Possible explanations for these ‘‘false-positive’’ results include an unde899
BRIEF REPORT Table 3. Scoring algorithm for identifying NPM1 mutated patients based on % dissimilar staining (at threshold 1.1) and mean fluorescence intensity SCORE
% Dissimilar staining 51% dissimilar 20–50% dissimilar \20% dissimilar Mean fluorescence intensity (MFI) [4,500 \4,500 Total score Potential NPM1 positivea Negative* a
2 1 0 1 0 2 \2
Results to be confirmed by routine molecular testing.
systematic review of the digital images to interpret the cellular localization of NPM over statistical analysis alone. The imaging flow cytometry method we describe is a novel approach to qualitative and quantitative assessment of cellular NPM in AML. The addition of digital images to standard quantitative and statistical measurements makes this the most sensitive flow cytometric method available for the assessment of cellular NPM localization. With the incorporation of a ‘‘scoring’’ system using both the MFI and dissimilar staining %, all cases of NPM1-mutated AML were identified. However, although ‘‘false-positive’’ cases were seen and despite the novel methodology used by this approach, we believe that imaging flow cytometry has a place as a first-line technique to identify NPM1-mutated patients, but that it should be complimentary to current gold-standard molecular genetic assessment of NPM1 mutation status.
ACKNOWLEDGMENTS tectable non-exon 12 NPM1 mutation, NPM1 gene overexpression, interaction of other molecules or genes resulting in shuttling of NPM from the nucleus, other nuclear transport abnormalities, protein degradation, and protein structural changes (9,14). During our analysis, we also noted that the ImageStream was sensitive to intranuclear distribution of NPM staining, with increased heterogeneity resulting in higher dissimilar staining percentages in NPM1 wild-type AML cases. However, further analysis of our ‘‘false-negative/ positive’’ showed that this was not a contributing factor in this study. It is important to note that the similarity feature is a staining intensity-independent feature, and therefore interpatient staining variability would not affect the results to the extent of creating ‘‘false-positives’’ and/or negatives. Although 49 patients were selected for the study, only 37 were assessed. Eight samples were excluded preanalysis because of poor quality and/or low-cell numbers, but an additional four samples were also excluded postanalysis due to poor/weak NPM staining despite technical standardization. It is of note that the poor/weak staining was confined to NPM1 wild-type cases. This supports the previous data that NPM MFI in NPM1 wild-type AML is lower than mutated cases and that these cases may represent extremely low-NPM levels [10]. The similarity feature within the IDEAS software is based upon the log-transformed Pearson’s correlation coefficient, measuring the degree to which two images are linearly correlated within a masked region, that is, the nucleus and/or cytoplasm. Therefore, in the event of the nuclear protein translocating to the cytoplasm, less protein stain is detected in the nuclear area, and therefore the nuclear images become anticorrelated and ‘‘dissimilar.’’ In most of the AML cases with the NPM1 mutation, the majority of NPM protein fluorescence was localized within the cytoplasm with minimal residual nuclear stain, thereby resulting in negative correlation and correct classification. However, the amount of cytoplasmic and residual nuclear NPM was variable. Residual nuclear fluorescence was seen in some cases and may have affected the algorithm by giving a more ‘‘similar’’ result despite significant detectable cNPM fluorescence on imaging. This highlights the added value from 900
The authors thank David Basiji (CEO, Amnis Corporation) for his technical advice and constructive and critical review of the manuscript. Graham Bottley worked as a technical consultant for Cronus Technologies who marketed the Amnis ImageStream in the UK during this research project. No other conflicts of interest are declared.
LITERATURE CITED 1. Cordell J, Pulford K, Bigerna B, Roncador G, Banham A, Colombo E, Pelicci P, Mason D, Falini B. Detection of normal and chimeric nucleophosmin in human cells. Blood 1999;93:632–642. 2. Dumbar TS, Gentry GA, Olson MO. Interaction of nucleolar phosphoprotein B23 with nucleic acids. Biochemistry 1989;28:9495–9501. 3. Falini B, Mecucci C, Tiacci E, Alcalay M, Rosati R, Pasqualucci L, La Starza R, Diverio D, Colombo E, Santucci A, et al. Cytoplasmic nucleophosmin in acute myelogenous leukemia with a normal karyotype. N Engl J Med 2005;352:254–266. 4. Arber D, Brunning R, LeBeau M, Falini B, Vardiman J, Porwit A, Thiele J, Bloomfield C. Acute myeloid leukaemia with recurrent genetic abnormalities. In: Swerdlow SH, Campo E, Harris NL, Jaffe ES, Pileri S, Stein H, Thiele J, Vardiman JW, editors. WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues, 4th ed. Geneva, Switzerland: IARC Press; 2008; p 110–123. 5. Falini B, Bolli N, Bhan J, Martelli M, Liso A, Pucciarini A, Bigerna B, Pasqualucci L, Mannucci R, Rosati R, et al. Both carboxy-terminus NES motif and mutated tryptophan(s) are crucial for aberrant nuclear export of nucleophosmin leukemic mutants in NPMc1 AML. Blood 2006;107:4514–4523. 6. Falini B, Martelli M, Bolli N, Bonasso R, Ghia E, Pallotta M, Diverio D, Nicoletti I, Pacini R, Tabarrini A, et al. Immunohistochemistry predicts nucleophosmin (NPM) mutations in acute myeloid leukemia. Blood 2006;108:1999–2005. 7. Mattsson G, Turner SH, Cordell J, Ferguson DJ, Schuh A, Grimwade LF, Bench AJ, Weinberg OK, Marafioti T, George TI, et al. Can cytoplasmic nucleophosmin be detected by immunocytochemical staining of cell smears in acute myeloid leukemia? Haematologica 2010;95:670–673. 8. Gruszka AM, Martinelli C, Sparacio E, Pelicci PG, de Marco A. The concurrent use of N- and C-terminal antibodies anti-nucleophosmin 1 in immunofluorescence experiments allows for precise assessment of its subcellular localisation in acute myeloid leukaemia patients. Leukemia 2011;26:159–162. 9. Gruszka AM, Lavorgna S, Consalvo MI, Ottone T, Martinelli C, Cinquanta M, Ossolengo G, Pruneri G, Buccisano F, Divona M, et al. A monoclonal antibody against mutated nucleophosmin 1 for the molecular diagnosis of acute myeloid leukemias. Blood 2010;116:2096–2102. 10. Oelschlaegel U, Koch S, Mohr B, Schaich M, Falini B, Ehninger G, Thiede C. Rapid flow cytometric detection of aberrant cytoplasmic localization of nucleophosmin (NPMc) indicating mutant NPM1 gene in acute myeloid leukemia. Leukemia 2010;24:1813–1816. 11. Grimwade L, Gudgin E, Bloxham D, Scott MA, Erber WN. PML protein analysis using imaging flow cytometry. J Clin Pathol 2011;64:447–450. 12. Mirabelli P, Scalia G, Pascariello C, D’Alessio F, Mariotti E, Noto RD, George TC, Kong R, Venkatachalam V, Basiji D, et al. Imagestream promyelocytic leukemia protein immunolocalization: In search of promyelocytic leukemia cells. Cytometry A 2012;81:232–237. 13. Tan AY, Westerman DA, Carney DA, Seymour JF, Juneja S, Dobrovic A. Detection of NPM1 exon 12 mutations and FLT3—Internal tandem duplications by high resolution melting analysis in normal karyotype acute myeloid leukemia. J Hematol Oncol 2008;1:10. 14. Vassiliou GS, Cooper JL, Rad R, Li J, Rice S, Uren A, Rad L, Ellis P, Andrews R, Banerjee R, et al. Mutant nucleophosmin and cooperating pathways drive leukemia initiation and progression in mice. Nat Genet 2011;43:470–475.
Detection of cNPM Expression by Imaging Flow Cytometry
