The effect of erythrocyte lysing reagents on enumeration of leukocyte subpopulations compared with a no‐lyse‐no‐wash protocol

Standard protocols in flow cytometry (FCM) require lysis of erythrocytes, which may induce an unwanted loss of leukocytes as bystander effect.


| INTRODUC TI ON
The accurate enumeration of leukocyte subsets is of clinical relevance. As an example, CD4 counts are of interest for HIV patients due to the appropriate start of medical intervention. 1 Immunophenotyping by FCM has been reported to be essential in this regard, and therefore, FCM analysis has become an integral part in this field of hematologic diagnostics.
A number of different strategies for sample preparation can be incorporated in the staining procedure before readout in the cytometer. Whereas cell separation using Ficoll is still frequently used for qualitative analysis of blood cells, almost all protocols for quantitative determination of blood cell subsets require a lysing step before analysis in order to remove erythrocytes. Although it is known that the lysing procedure may lead to distinct loss of cells as it is described previously for CD4+ cells, 2 CD34+ counts, [3][4][5] and total leukocyte counts. 5 Further effects like alteration in morphology, light scatter properties, antigenic epitopes of leukocytes and thus definition of subpopulations have also been described. 6 Theoretically, the only way to avoid these undesirable effects is to completely eliminate lysing steps by applying no-lyse-no-wash protocols. [7][8][9] However, due to the very high load of flow cytometer fluidics with unlysed erythrocytes, other problems may arise with the latter approach.
In the present study, we systematically compared completeness of RBC lysis and effects of different lysis reagents on major leukocyte population counts as well as lymphocyte subpopulations. We investigated the influence of four lysing reagents and six different strategies in total (five lyse-no-wash strategies and one no-lyseno-wash-strategy) on the enumeration of leukocyte subsets. We analyzed more leukocyte subsets compared to earlier studies, using antibodies against CD3, CD4, CD8, CD19, CD14, CD16, CD56, and CD45. We analyzed if different leukocyte subsets show different sensitivities to different lysing reagents, which has to be considered for each individual case when specific subsets are of interest.
Moreover, we tested if a no-lyse-no-wash protocol may be a feasible alternative to lyse-no-wash protocols.

| Samples and reagents
We obtained 52 blood samples as part of the routine diagnostics in our department from February to May 2015. Whole blood was collected into EDTA containing tubes and processed within 4 hours of acquisition. Only samples with 2-10 × 10 6 leukocytes/µL and a roughly physiological distribution of leukocyte populations were included in this analysis (granulocytes: 40%-80% of all leukocytes; lymphocytes: 20%-50% of leukocytes; monocytes: 3%-15% of leukocytes), because systematic comparisons would not be possible if populations were completely missing in pathological samples. For control purposes, we routinely prepared one blood smear per sample for microscopic analysis. IMMUNOTECH SAS, Beckman Coulter) (VersaFix). Five aliquots per sample were analyzed by lyse-no-wash protocols using the reagents described above and one by a no-lyse-no-wash protocol (NoL). The following antibodies were used: CD16-FITC, CD56-PE, CD3-ECD, CD4-PC7, CD19-APC, CD14-APC700, CD8-PB, and CD45-KO (Beckman Coulter).
Thereupon, the 6 aliquots of stained sample were distributed into different tubes and processed according to manufacturers' instructions as described below. All procedures were carried out at room temperature.

| Data analysis and gating
Data analysis was performed using Kaluza software (Beckman Coulter). FS and SS were used as untransformed linear values; fluorescence channels were transformed to a logical scale. The gating strategy used for identifying different cells is summarized in Table 1 and shown in Figure 1.
We set three gates, FS/SS, TIME/SS, and SS/CD45, to include all leukocytes ("Leuko") but exclude erythrocytes and debris. The The second, stricter gating strategy ignores the minor subpopulation of CD14 negative monocytes, but is less prone to technical errors such as overlap with the granulocyte gate or B-cell gate. Gating strategies for all populations are depicted in Figure 1.
The NoL method caused a partial shift of lymphocytes into areas of higher SS. Therefore, we include also lymphocytes with high SS in these samples. The FS INT/FS PEAK dot plot served as quality control to rule out the existence of significant numbers of doublets.

| Analysis of population sizes
Completeness of RBC lysis was calculated as a fraction of Leuko events of all events collected excluding FS low events by an acquisition threshold as described above. We calculated population sizes as fractions of all leukocytes as defined above. We performed control calculations to make sure that any cells are neither omitted nor double-counted and to have an internal quality control for our gating strategy: Granulocytes + Mono CD14 + Lymphocytes = L eukocytes; B cells + T cells + NK cells = Lymphocytes; T-helper (Th) + Cytotoxic T (Tc) + double negative T (Tdn) = T cells. We accepted deviations of less than 1% for lymphocytes and T cells and less than 2% for Leuko control calculations. If variations exceeded this limit, we re-checked and adjusted gates if necessary. The socalled "Bermuda triangle" or "blast region," defined as CD45 dim/SS low, containing no positively defined cells (ie, blasts and basophils) was part of the Leuko gate. This was specifically checked if the internal control calculation for leukocytes was below 98%.
Considering the leukocyte event counts of 30 000, we defined a lower limit of 0.2% (ie, a minimal cell count of at least 60 cells) for any subpopulation to be included in our comparisons. Populations below this size were not used for the comparison of lysis procedures.
This criterion was only relevant in a few samples for the following lymphocyte subsets: B cells, T-CD56+, and Tdn.
Since population percentages showed a wide variation in different individuals, population sizes were analyzed as relative values compared to the reference size for this population. For example if the NoL preparation contained 12% T cells, this was defined as reference = one. Our predefined reference was the NoL method under the assumption that this method would provide results closest to "true" values in vivo. Since the definition of population size was problematic in some samples using the NoL method due to the shift of FS and/or SS, we used the mean value of all methods as additional reference for the "true" population size.

| RE SULTS
We tried to eliminate possible technical gating ambiguity by setting robust gates. Nevertheless, separation of some populations caused problems, mainly after NoL regarding the FS and SS plot, that is, the NoL method led to an unexpectedly wide distribution of FS and SS ( Figure S1). For a large part, this may be due to transient doublet formation in the erythrocyte-rich cell suspension, because FS and SS decreased after application of a doublet discrimination gate. Therefore, we include also lymphocytes with high SS signals in measurements of no-lyse samples. Since this might interfere with our attempt to use NoL as reference, we always checked the mean of all methods as a second reference method. However, as predefined in our analysis plan, all primary results below refer to the NoL method as a reference. Results using the methods mean led to similar results (see below).

F I G U R E 1
Flow cytometric gating strategy for determination of leukocytes and subsets in a representative blood sample. The sequence of color coding and logical combination of gates is depicted in

SS differed between preparations and was increased in FacsL
treated samples. This phenomenon was very prominent for eosinophil granulocytes (granulocytes after subtraction of neutrophils, that is, SS high/CD16-cells with CD45 slightly above the neutrophil population). These cells are shifted above the upper SS measurement range in the FacsL preparations and thus seem to be missing from the plots including SS ( Figure S1), but are displayed in other plots as separate population with high background staining (eg, CD3/CD8 plot in Figure 1).
Two measurements had to be discarded from the analysis because of failed preparation or individual cell count abnormalities. We decided to include 12 samples only very slightly deviating from our physiological distribution preconditions defined above.
Control calculations "Sum of Leuko" were below 98% in 5 measurements. In all 5 cases, missing cells clustered in the CD45 dim/SS low "Bermuda region," thus are probably basophils and/or blast cells and biologically plausible.

| Completeness of RBC lysis
Lysing efficiency was significantly different between the lysis methods (P < .0001). Most efficient lysis of RBC was achieved by VersaL (median Leuko 96.92% of all events) and NH4Cl (96.85%), closely followed by VersaFix (93.55%). In contrast, we observed less reliable RBC lysis with QuickL (79.05%) and FacsL (59.06%), both of the latter methods significantly inferior compared to each of the first three methods (P < .0001 for all comparisons, Figure 2).

| Major leukocyte populations
Influence of the lysing strategy upon enumeration of leukocyte population sizes was variable in different samples with a broad overlap and a few outliers. However, some differences were clearly significant. Figure 3 shows the relative population sizes as a percentage of  Figure S2), with the exception of the monocyte fraction, which was measured somewhat smaller with NoL compared to the average of all methods. Again, similar results were obtained using the mean of all methods as reference ( Figure S3). However, almost all well-established protocols require an erythrocyte lysing step during sample preparation, which still may lead to lysis-induced cell loss. Therefore, some authors suggest no-lyse protocols in order to avoid inaccurate counts. [7][8][9] The aim of this study was to expand investigations of effects of lyses reagents on more different leukocyte subsets using lysis reagents that are frequently used nowadays. In addition, we compared the completeness of RBC lysis.

F I G U R E 4
Effect of erythrocyte lysing reagents on the enumeration of lymphocyte subsets. Size of lymphocyte subsets (B, T, Tc, Th, Tdn, T-CD56+, and NK) as fraction of total lymphocytes was calculated and is shown here as relative value in comparison with the fraction size obtained by the NoL method (reference, defined as 1) We found best results in completeness of RBC lysis after treatment with NH4Cl and VersaL, closely followed by VersaFix. In contrast, significantly inferior results with high variability were observed with QuickL or FacsL.
In contrast, Bossuyt et al defined debris after setting a very low forward scatter acquisition threshold, which resulted in very high amounts of debris using NH4Cl and most other reagents. 11 However, this strategy is in contrast to common everyday practice to set the forward scatter threshold just sufficiently below the size of the smallest cells expected to be acquired.
Regarding the main leukocyte populations, we found that granulocyte and neutrophil percentages were higher after FacsL compared to unlysed samples. We observed slightly higher numbers of monocytes compared to NoL with all reagents except QuickL. Using the latter, we obtained lower monocyte counts compared to NoL (−11%). The lymphocyte fraction was estimated 19% lower after FacsL. Thus, we observed meaningful differences between major cell populations. In contrast, differences within the lymphocyte compartment were not very pronounced. However, with FacsL, we counted larger fractions of T-CD56+-cells and NK cells and lower fractions of B cells.
Only few publications report data that can be compared directly to the results presented here. Many reports deal only with progenitor cells or blasts, [3][4][5]7,9,10,12 others describe only effects of lysing reagents that were not part of our study and/or are no longer commonly used. [2][3][4][5]9,13,14 Different counts of leukocyte subsets after different lysing procedures can either be the result of real differences in cell losses during the procedure or due to shifts in scatter properties or staining intensities leading to different gating and thus only mimicking differences in the composition of cell subsets. As others did before, 6,11,13,15 we also observed modifications in scatter properties after different lysis treatments. Most of these differences could be compensated by adjustments of FS and SS detector sensitivities. Nevertheless, positions of main populations remained slightly different. FacsL resulted in inferior discrimination of granulocytes from monocytes in SS/CD45 plots, but better discrimination of monocytes from lymphocytes. It was described by others that FacsL and ammonium chloride-based lysing reagent yielded the best discrimination via scatter prosperities in the FS/SS plot using Becton Dickinson cytometers. 15  prolonged. Furthermore, light scatter properties are changed which complicates cell discrimination in the FS and SS channels. To cope with this phenomenon, we applied a gating strategy including a very wide distribution of SS signals in our study. Application of an FS peak vs. integral gate to exclude doublets is undesirable, because this would also lead to a loss of stained cells, that is, leukocyteerythrocyte doublets. Taken together, the NoL method as a reference confirmed the results from three of the lysis methods tested, but we would not recommend this strategy as a routine procedure in everyday practice. This holds true especially, if antibody panels are to be used, which strongly depend on the analysis of FS and/or SS for gating purposes.
In summary, our data show that the lysing reagents tested lead to specific deviations in the quantitation of leukocyte subsets and show different efficiency of erythrocyte lysis. Thus, intra-individual follow-up measurements should preferably be performed with the same antibodies, lysis procedure, and instrument settings. Three strategies (NH4Cl, VersaL, and VersaFix) using two different reagents are in good accordance with each other and with a NoL reference method. Quantitation of subpopulations within the lymphocyte compartment is quite similar among all methods. NoL is a feasible alternative to lysis strategies, but entails other disadvantages due to slower acquisition and altered FS and SS signals.

ACK N OWLED G EM ENTS
The author would like to thank the technicians of the hematologic laboratory of Medizinische Klinik 5, University Hospital Erlangen, for their excellent assistance with measurements and data collection. We thank Cytognos for providing a sample of QUICKLYSIS™ for study purposes. The present work was performed in fulfillment of the requirements of obtaining the degree "Dr med." for KP

CO N FLI C T O F I NTE R E S T
The authors declare no potential conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Raw data are available on request due to privacy/ethical restrictions.