Evaluation of bone marrow aspirates using the automated hematology analyzer Sysmex XN‐3000

This study evaluated the feasibility of the Sysmex XN‐3000 automated hematology analyzer for the assessment of total nucleated cells (TNC) and bone marrow (BM) cell density in routine bone marrow aspiration (BMA) samples.


| INTRODUCTION
Bone marrow (BM) examinations using bone marrow biopsy (BMB) and bone marrow aspiration (BMA) are essential in diagnosis, evaluation of prognosis, and therapeutic responses for a variety of hematological disorders. Microscopic examinations of BMBs give semi-quantitative BM cellularity which is indispensable for clinical diagnoses as it determines the ratio of total nucleated cells (TNCs) and lipid particles. However, when biopsy cores are not available and/or quick assessment is necessary, BM cellularity is estimated from the ratios of hematopoietic and adipocytic areas observed on particle crush smear slides of BMAs along with manual counting of TNCs with a hemocytometer. [1][2][3] Megakaryocytes in BMAs are also counted using a hemocytometer because megakaryocyte counting is important in differentiating the underlying causes of thrombocytopenic platelet disorders with normal platelet morphology. These disorders are caused by either decreased platelet production or increased platelet destruction, with decreased and increased megakaryocytes in BM, respectively. 4 These BM examinations have been conventionally performed by manual microscopic methods, which is tedious and imprecise because it depends heavily on the experience of the technologists. Therefore, the potential advantages of BMA analysis using an automated hematology analyzer have been noted to improve efficiency, throughput, reproducibility, objectivity of measurement, and also decreases manual labor, costs, and variability. [5][6][7][8][9][10] Though the feasibility of automated hematology analyzers to examine BMA samples have been studied, 5-10 the accuracy of automated analyzers have been shown to be inferior to manual microscopic examinations. [5][6][7]11 This is mainly due to the complexity of cell compositions in BMs compared to peripheral blood cells. 3,8 In addition, automated analyses of BMA samples are hampered by contaminations such as lipid droplets, microfibers, and solid particles. 5 By using BMAs obtained from patients with various hematological disorders, we sought to investigate the feasibility and accuracy of the Sysmex XN-3000 analyzer and found promising potential in assessing BM cellularity by measuring the ratios of TNCs and lipid particles.
We also demonstrated the limitations of using the number of megakaryocytes detected by the XN-3000 as a direct quantitative count of megakaryocytes.

| Patients
Fifty-four BMA samples were obtained from 39 patients for diagnosis and follow up of hematological disorders at Juntendo University Hospital between May 2017 and August 2019. Table 1 lists the details of the diagnoses. This study was approved by the Juntendo University Hospital Medical Ethics Committee (Tokyo, Japan), and was conducted in strict adherence to the guidelines by the Declaration of Helsinki. As part of the approval, the Internal Review Boards at Juntendo University Hospital explicitly waived the need for written informed consents from the patients as all the samples were de-identified.

| Cell counting and differentiation by the XN-series
Differential counting was carried out with Sysmex XN-3000 (Sysmex Corporation, Kobe, Japan), and its performance was assessed by comparing manual cell counts and differential results.
BM samples were collected from the patients' posterior iliac crest and were stored in tubes containing K 2 -EDTA. The samples were then filtered to remove some sources of interference such as bone fragments and blood clots. The size of megakaryocytes, which are normally $37 μm in diameter, has been shown to increase to an average of 46 μm (range, 21-56 μm) in essential thrombocythemia (ET), a myeloproliferative neoplasm (MPN) with megakaryocyte hyperproliferation. 12 Therefore, we used a 190 μm pore filter that allows larger megakaryocytes in ET cases to be safely retrieved. Within 1-2 h after collecting the samples, numbers of TNCs, nucleated red blood cells (NRBCs), and white blood cells (WBCs) were measured by the XN-3000 equipped with WNR and WBC differential (WDF) channels. These channels simultaneously generate three different signals from each cell passing through the focused laser beam in the detecting chamber known as Flowcell: (1) forward scatter signals indicating cell volume; (2) side scatter signals reflecting intracellular structures and contents, such as nuclei and granules; and (3) fluorescence intensity signals indicating amounts of intracellular nucleic acids (i.e., DNA and RNA). 13 Based on these signals, number and differentiation of the cells were analyzed and categorized, and the results were shown as "scattergrams." The WNR channel was used to differentiate WBC, NRBC, and basophil (BASO) counts, whereas the WDF channel was used to classify neutrophils, lymphocytes, monocytes, eosinophils, and immature granulocytes (IGs). Both channels use flow cytometry with a semiconductor laser. [13][14][15] Signal intensities were indicated with T A B L E 1 Diagnosis of the patients.

Diagnosis Number
Acute myeloid leukemia 5 Acute promyelocytic leukemia 5  Because megakaryocytes with large size and high fluorescence intensity appear outside the region of WBCs, the number of megakaryocytes was counted using the WDF channel of the body fluid mode of XN-3000 (BF-XN). 19 Using the Flowing Software, megakaryocytes were distinguished from WBCs in the scattergram using sidescattered light (x-axis) and side fluorescent light (y-axis). The region in which fluorescence intensity is higher than that of the WBC region, which is higher than 225 channels in the side fluorescent light, was gated. The number of dots were then counted.

| Manual counting and differential
To determine TNC and megakaryocyte counts as a reference, manual microscopic counting, commonly performed in Japanese hematology laboratories, was performed according to laboratory protocols. [20][21][22][23] Briefly, 20 μL of BMA sample was diluted with 180 μL of Türk solution, then the numbers of TNCs and megakaryocytes were manually counted under the microscope using Bürker-Türk and Fuchs-Rosenthal hemocytometers (Thermo Fisher Scientific, AL). Each sample was counted four times and the average count was calculated.
Number of TNCs was expressed as the number of nucleated cells per microliter of BM, which is comparative to the density unit of XN.
Manual differential counting was performed on May-Grunwald

| BM biopsy specimen preparation and microscopy examinations
Lipid contents measured by XN-3000 were compared with BM cellular density evaluated by BMBs. Pathological examinations were performed at the pathology laboratory in Juntendo University Hospital.
BMB samples were prepared based on the guidelines published by the ICSH. 25 BMB specimens were processed with fixation and decalcification. After decalcification, the specimens were embedded in paraffin wax, and sections were cut using a microtome. The biopsy sections were stained with hematoxylin and eosin. BM cellularity was evaluated as volume ratios of hematopoiesis and lipids. The hematopoietic cellularity was defined as the proportion of cellular elements relative to marrow adipose tissues by hematopathology, 26 then the sum of the hematopoietic and adipocytic areas as the denominator for the calculation of "cellularity". Equation (1) is written as 27 : BM cellularity was defined as the following: ≥70%, hypercellular; 30%-70%, normocellular; and ≤30% hypocellular. 28   We evaluated the accuracy of TNC counting using the WNR channel of XN-3000. Figure 1A shows a representative scattergram of TNCs consisting of WBCs, NRBCs, and BASOs. The BM specific plasma cells and the reticuloendothelial cells were also included in the counts of TNC. The TNC counts by XN-3000 were compared with the manual microscopic examinations ( Figure 1B) showing that the TNC counts by XN-3000 were linearly correlated with the counts by manual microscopic examinations (R 2 = .98, p < .001). These findings were confirmed by the Passing-Bablok regression (r = .66, 95% CI 0.17-0.89) and Bland-Altman plot method ( Figure S1).

| Statistical analysis
To count lipid particles, we utilized the WDF channels. Figure 2A shows a representative scattergram of a sample including lipid particles with sigmoidal-shaped distribution (blue dots) that overlaid with WBCs (forward and side scattered light signals). Figure 2B shows the lipid particles separated from the nucleated WBCs (side fluorescent and side scattered light signals). After counting, the nucleic acid-free lipid particles and RBC debris, which were detected below the signal intensity of 15 channels, 17,18 were re-gated (forward-and side-scattered light signals) ( Figure 2B). To exclude small RBC debris, the number of dots located higher than 30 channels of the forward-scattered light were counted as lipid particles ( Figure 2C). The sensitivity and specificity for hypoplasia were 100% and 88%, respectively; and the sensitivity and specificity for hyperplasia were 89% and 86%, respectively.

| Counting of megakaryocytes by XN-BF mode
Another challenge in analyzing BMA samples using the automated hematology analyzer is counting large megakaryocytes. The megakaryocytes with large and high fluorescence intensity were measured by the gating algorithm using the WDF channel of XN-BF mode as described in the Section 2. The number of dots in the region where the fluorescence intensity is higher than the WBC region (225 channels) in the side fluorescent light was counted ( Figure 3A). Figure 3B shows a linear correlation of megakaryocyte counts between the XN-3000 and the manual microscopic examinations (R 2 = .59, p = .002). The correlation was also evaluated by Passing-Bablok regression (r = .66, 95% CI 0.17-0.89). The Bland-Altman plot demonstrated that the value of one sample from chronic myelogenous leukemia was lower than À2SD ( Figure S2).

| Analysis of NRBC and M/E ratio in BM aspirates by XN-3000
The gating maps for the NRBC and WBC differential of BM aspirates were constructed based on the cytograms analyzed using the WNR The number of lipid particles was measured by the gating algorithm using the WDF channel as described in the Section 2. The ratio of lipid particle numbers/TNC numbers measured by XN-3000 was calculated as the index of nucleated cells density in BMA. *p < .001. (E) The measurement of BM cellularity as a control method, the ratio of hematopoietic cells to lipid, was determined with H&E-stained sections of the BM biopsy in the pathology laboratory. (F) The measurement of BM cellularity as a control method, the proportion of cellular elements relative to marrow adipose tissue, was determined with H&E-stained sections of the BM biopsy in the pathology laboratory. *BM cellularity was calculated as described in the Section 2; % of hematopoietic area/(hematopoietic area + adipocytic area). ≥70%, hypercellular, n = 11; 30%-70%, normocellular, n = 10; and ≤30% hypocellular, n = 7. (G) Receiver-operating characteristic (ROC) curves for prediction of hypoplasia or hyperplasia using XN-3000 parameters regarding the numbers of lipid particle and TNC.

| DISCUSSION
In this study, we assessed the BM nuclear cell density by counting the lipid particles of BMAs, counted the numbers of TNC and megakaryocytes, and determined the M/E ratios using Sysmex XN-3000. Using the newly equipped fluorescence detectors, WNR channels, we successfully gated the lipid particles, and then differentiated the lipid particles and WBCs. 18 The counted numbers of lipid particles were further utilized to estimate BM nucleated cell density. As a result, a good correlation was observed between the XN-3000 and the pathologist's scoring of cellularity with BMB using the estimated hematopoietic cellularity calculated by the numbers of TNC and lipid particles detected. Blue dots indicate samples with an M/E ratio of 4 or greater as measured by the microscope (n = 18, including chronic myelogenous leukemia, chronic myelomonocytic leukemia, atypical chronic myeloid leukemia, essential thrombocytosis, myelodysplastic syndromes, acute myeloid leukemia, acute lymphoblastic leukemia, B cell lymphoma, and reactive thrombocytosis). This is the first report of a fully automated hematology analyzer that successfully estimated BM nucleated cell density by measuring nucleated cells and lipid particles without affecting each other.
The gating method of the nucleic acid-free lipid particles was simple and considered to be feasible for practical use.
We further demonstrated that the plot corresponding to megakaryocytes appeared in a unique region of the scattergram using the WDF channel of BF mode that allows the XN-3000 to detect large cells. Though the XN-BF mode does not have a specific software to measure megakaryocytes, the simple gating algorithm was effective in counting the number of megakaryocytes. In addition, we con- BMA is a mixture of BM and sinusoidal blood which may not provide a reliable estimate of BM cellularity. 29 Therefore, confirmation by BMB, which allows qualitative assessment of nucleated cell density, is indispensable. 30,31 A previous study demonstrated that BMB successfully confirmed the interpretation of BMA smears in aplastic anemia by revealing hypocellular BM with focal hypercellular areas.
Another previous study demonstrated that BMB successfully confirmed the BMA smear interpretation of aplastic anemia by revealing focal hypercellular areas. 32 Although estimation of the BM cell composition still relies on the assessment of hematopoietic cellularity by low-magnification examinations of the BMB, several approaches to quantify BM lipid compartments along with hematopoietic cellularity have recently been reported as digital pathology workflows in tissue sections. [33][34][35] In the field of clinical and experimental pathophysiology, the image-analysis programs emerging for quantification and detection of specific cell types related to disease progression are developing. 36,37 These morphological machine learning methods attempt to quantitatively evaluate overall BM architecture in H&E-stained sections by subclassification of the bone, adipose, hematopoietic, and interstitial/microvascular compartments. However, they still have to determine the distinct morphology of specific cells, including megakaryocytes with their large sized cytoplasm and the extent of the artifact of the samples.
To increase accuracy, comparability, and reproducibility of the results, automatic measurement and quantification is required. 34 The methodology of this study successfully provided quantification of hematopoietic cellularity with lipid particles in BMA samples utilizing an automated hematology analyzing system.
There are several limitations in this study. First, this is a single-

ACKNOWLEDGMENTS
We would like to sincerely express our gratitude to, Akihiko Matsuzaki, Kumiko Nishibe, Setsuko Marutani, Maiko Yuri, and Yoshie Hosaka for their technical assistance on this project.

FUNDING INFORMATION
This work was supported in part by Japan Society for the Promotion of Science Grants-in Aid for Scientific Research (20K20230 to Yuki Horiuchi and 22H02974 to Yoko Tabe).