B L O O D C O M P O N E N T S MEASUREMENT OF CELL VIABILITY

HPC viability measurement: trypan blue versus orange and

K. Mascotti, J. McCullough, and S.R. Burger

reliable method for rapidly determining the vi- BACKGROUND: A reliable, validated method for rapidly ability of HPCs is essential for clinical cellular en- determining HPC viability is essential for clinical cell en- gineering. The trypan blue (TB) viability assay is gineering. a widely performed procedure to assess HPC vi- STUDY DESIGN AND METHODS: A fluorometric cell Aability before and after cryopreservation and before bone viability assay using acridine orange and propidium io- marrow transfusion.1 The TB assay is based on the ability dide (AO/PI) was compared to the current standard, of the membrane of viable cells to exclude the dye, while trypan blue (TB) exclusion. Viable cells stained with AO/ nonviable cells are stained blue when viewed by light mi- PI fluoresce green under darkfield micros- croscopy. However, the TB assay can be difficult to interpret copy, while nonviable cells fluoresce orange. Mixtures of because of artifacts. In addition, the TB viability fresh and heat-killed bone marrow were prepared and assay has never been validated for use with HPCs. used as viability standards for evaluation of both assays. Alternative viability assays using the fluorescent dyes The frequency of CFU–GM was determined for each acridine orange (AO) and propidium iodide (PI) have been specimen. developed for the simultaneous visualization of both viable RESULTS: Cell viability measured by AO/PI was ex- and nonviable cells.2,3 AO is a membrane-permeable, cat- tremely linear, with measured and predicted viability in ionic dye that binds to nucleic acids of viable cells and that agreement from 0 to 100 percent of the viable cells and at low concentrations causes a green fluorescence. PI is a coefficient of regression (r2) of 0.9921. The predicted- impermeable to intact membranes but readily penetrates viability regression line fell within the 95% CI for AO/PI- the membranes of nonviable cells and binds to DNA or measured viability. The coefficient of regression for TB- RNA, causing orange fluorescence. When AO and PI are measured viability was 0.9584, with the predicted-viability used simultaneously, viable cells fluoresce green and non- regression line almost entirely outside the 95% CI. TB viable cells fluoresce orange under fluorescence micros- overestimated the percentage of viable cells, particularly copy. Results of the AO/PI assay in pancreatic islet cells have below the 50-percent level. CFU–GM frequency corre- been shown to correlate with those of the TB viability as- lated better with cell viability measured by AO/PI (r2 = say but have been easier to interpret.4 0.979) than with that measured by TB (r2 = 0.930). Here we describe a rapid fluorometric assay for HPC CONCLUSIONS: The AO/PI viability assay is a rapid, viability using AO/PI dyes and compare the results of this highly linear, functionally correlated assay that is supe- rior to conventional viability measurement by TB exclu- α sion. ABBREVIATIONS: AO = acridine orange; -MEM = minimum essential medium alpha modification; PI = propidium iodide; r2 = coefficient of regression; TB = trypan blue.

From the Department of Laboratory Medicine and Pathology, University of Minnesota Medical School, Minneapolis, Minne- sota. Address reprint requests to: Scott R. Burger, MD, Director, Cell Therapy Clinical Laboratory, University of Minnesota, Box 609, Mayo Building, 420 Delaware Street SE, Minneapolis, MN 55455; e-mail: [email protected]. Received for publication October 21, 1998; revision re- ceived August 23, 1999, and accepted September 2, 1999. TRANSFUSION 2000;40:693-696.

Volume 40, June 2000 TRANSFUSION 693 MASCOTTI ET AL.

assay with those of the TB viability assay. Both viability as- Progenitor assays were performed on each predicted- says were validated by the use of specimens of known vi- viability specimen by using serum-free methylcellulose ability and clonogenic assays for HPCs. In addition, the progenitor assay medium (MethoCult GF HC4434, Stem stability over time of the TB and AO/PI staining was determined. Cell Technologies, Vancouver, BC, Canada) and a standard- HPCs derived from peripheral or umbilical cord blood were ized progenitor assay method.6 One person performed all not tested. progenitor assays. Cells were suspended in the MethoCult medium and plated in duplicate 35-mm culture dishes at each of two plating concentrations, 2.5 × 104 and 5.0 × 104 MATERIALS AND METHODS nucleated cells per plate, with 1.1 mL dispensed per plate.

Cells Plates were incubated at 37°C in a humidified 5-percent CO2 Bone marrow was collected from the posterior iliac crest of incubator for 14 days. CFU–GM were counted with a ster- a normal donor under institutional review board approval eomicroscope (SZH-ILLD, Olympus Precision Instruments, and by using standard techniques.5 Fresh, heparinized bone Melville, NY) at 40× power. The plating concentration giv- marrow was washed into minimum essential medium al- ing optimal growth and readability was selected for each pha modification (α-MEM, GIBCO, Grand Island, NY) with pair of duplicate plates, and the mean concentration of 5 1-percent HSA (Baxter-Hyland, Glendale, CA). The cell sus- CFU–GM per 10 nucleated cells was calculated. pension was divided into two equal aliquots of 10 mL. Cells The stability of the TB and AO/PI viability assays over in one aliquot were heat-killed at 65°C × 15 minutes and time was determined by repeated measurement of the per- then washed and resuspended in a total volume of 10 mL centage of viable cells in a single, fresh marrow specimen of α-MEM with 1-percent HSA. These two aliquots were over 120 minutes. Two cell suspensions were stained with measured by both the AO/PI and TB assays; the cells in the TB and AO/PI, respectively, by the methods described heat-killed aliquot were 100-percent nonviable, and those above. The percentage of viable cells was measured at 10- in the other aliquot were 100-percent viable. Cell concen- minute intervals over 60 minutes and at 15-minute inter- vals over an additional 60 minutes. trations were measured in each aliquot and adjusted to an identical value, 9.4 × 106 cells per mL. Statistical methods Calibrated automated electronic pipettors (Rainin Elec- tronics, Woburn, MA) were used to prepare seven 1-mL cell Software programs (Excel for Windows 95, version 7.0, Microsoft, suspensions from these two stock aliquots. The suspen- Redmond, WA; SigmaPlot for Windows, version 3.06, SPSS , Chi- sions represented predicted percentages (0, 5, 25, 50, 75, 95, cago, IL) were used for data analysis. Viability measure- and 100%) of viable cells. Cell viability for each of these ments performed by each method were analyzed by linear seven aliquots was evaluated by both TB exclusion and AO/ regression, regression coefficients were compared with the ideal values of 0 and 1, and 95% CIs were determined. PI staining. The percentage of viable cells was calculated as the average of five viability measurements, each of which re- quired the scoring of 100 cells. Measurements were repeated RESULTS five times to assess the reproducibility of viability measure- The percentage of viable cells measured by TB exclusion for ment. Measurements were performed by two individuals. the controlled-viability specimens is depicted in Fig. 1. The equation for the TB measured-viability regression line was Assays y = 0.796× + 23.437, TB viability measurement was performed by standard with a coefficient of regression (r2) of 0.9584. The slope methods.1 TB solution (0.4% wt/vol, GIBCO) was mixed (0.796) and y intercept (23.437) of this regression line were with each of the seven predicted-viability aliquots, using significantly different from the ideal values of 1 and 0. The 190 µL of stain to 10 µL of cell suspension. The suspension regression line for predicted cell viability was almost en- was loaded into a Neubauer hemocytometer and scored tirely outside the 95% CI for TB-measured viability. The CV with a Zeiss compound light microscope at 250×. Cells that for the five viability measurements performed on each stained blue were scored as nonviable. specimen ranged from 0 to 63.3 percent. TB consistently Cell viability was evaluated by AO/PI staining as previ- overestimated the percentage of viable cells, particularly at ously described by Banks.2,4 AO/PI stock solution was prepared cell viabilities below 50 percent. as 1 mM AO and 1 mM PI in PBS (Sigma Chemical Co., St. Louis, The percentage of viable cells measured with AO/PI is MO). Samples of each of the predicted-viability aliquots (10 µL) shown in Fig. 2. Viability measured by AO/PI was extremely were mixed with 190 µL of AO/PI stock solution. The suspension linear, with measured and predicted viability consistently was loaded into a Neubauer hemocytometer and scored in agreement for the entire range of viability measurement, with a Zeiss compound at 250×. from 0 to 100 percent. The linear regression equation for Cells fluorescing green were scored as viable. Cells fluorescing AO/PI measured viability was orange, either fully or partially, were scored as nonviable. y = 1.009× + 2.148,

694 TRANSFUSION Volume 40, June 2000 MEASUREMENT OF CELL VIABILITY CFU–GM per 100,000 cells Measured percentage of viable cells Percentage of viable cells measured

Predicted percentage of viable cells

Percentage of viable cells predicted Fig. 3. The percentage of viable cells measured by TB exclusion   Fig. 1. The percentage of viable cells measured by TB exclusion ( ) compared with the frequency of CFU–GM ( ). Mean ± 1 ( ) compared with the predicted percentage of viable cells (—). SD for n = 5 measurements and regression line are shown for Mean ± 1 SD for n = 5 measurements, regression line (– –), and both the percentage of viable cells (- - -) and CFU–GM (—). 95% CI (·····) are shown. CFU–GM per 100,000 cells Percentage of viable cells measured

Percentage of viable cells measured Predicted percentage of viable cells

Fig. 4. The percentage of viable cells measured by AO/PI fluo- Percentage of viable cells predicted rescence ( ) compared with the frequency of CFU–GM ( ). Fig. 2. The percentage of viable cells measured by AO/PI fluo- Mean ± 1 SD for n = 5 measurements and regression line are rescence ( ) compared with the predicted percentage of viable shown for both the percentage of viable cells (– –) and CFU– cells (—). Mean ± 1 SD for n = 5 measurements, regression line GM (—). (– –), and 95% CI (·····) are shown. dicted percentage of viability and was appropriately linear. The percentage of viable cells measured with TB correlated with r2 = 0.9921. Slope and intercept of this regression line moderately well with CFU–GM concentration (r2 = 0.930), were not significantly different from 1 and 0, respectively, but it consistently overestimated HPC frequency. The per- and the regression line for predicted viability fell within the centage of viable cells measured with AO/PI correlated 95% CI. The CV for the five viability measurements per- better with CFU–GM frequency (r2 = 0.979). formed on each specimen ranged from 1.3 to 68.8 percent. Figure 5 depicts the percentage of viable cells, mea- The results of both TB and AO/PI viability assays were sured over 2 hours by TB and AO/PI. The AO/PI results were compared with the frequency of CFU–GM (the actual num- more stable over time, consistently averaging 98.4 ± 0.67 ber of colonies grown from each of the two plating concen- percent over 120 minutes, with a linear regression equation trations), a measure of HPC function. Figure 3 depicts the slope of –0.0063, which is not significantly different from an percentage of viable cells measured with TB, compared with ideal value of 0. TB-measured viability, by contrast, de- the frequency of CFU–GM in the same samples. The per- creased from an average of 96.2 ± 1.60 percent during the centage of viable cells measured by AO/PI, compared with first 60 minutes, to an average of 89.8 ± 2.17 percent during CFU–GM frequency, is shown in Fig. 4. The measured CFU– the final 60 minutes with a linear regression equation slope GM frequency in each sample was proportionate to its pre- of –0.0862, which is significantly different from 0 (p<0.005).

Volume 40, June 2000 TRANSFUSION 695 MASCOTTI ET AL.

One reason that cell viability measured with AO/PI so closely correlated with predicted viabilities and CFU–GM concentration may be the assay’s greater readability and more accurate interpretation. With the high background of RBCs and debris as well as the variable staining or shading of cells, readability and scoring of cells with the TB assay can be ambiguous. In the AO/PI assay, RBCs and debris are not visible during darkfield microscopy, the cellular detail and nuclear outline are distinct, and the simultaneous use of two dyes allows for clear differentiation of viable and nonvi- Percentage of viable cells able cells. With AO/PI, therefore, there is no need to remove RBCs by lysis or ficoll density gradient centrifugation. Both assays are rapid and easy to perform, although the AO/PI assay requires a fluorescence microscope. We believe the in- Time (min) creased accuracy of results and ease of interpretation more than compensate for the need for fluorescence microscopy. Fig. 5. The percentage of viable cells repeatedly measured by TB is toxic to cells over time and affects the stability of AO/PI (– –) and TB (·· ··) assays over time allows comparison TB-stained cells.1 By contrast, the AO/PI assay has previ- of assay stability. Regression lines (—) for each assay are ously been described as being stable over time, and AO/PI shown. is nontoxic to isolated islet cells.2 A perfectly stable assay would have a linear regression slope of 0 over time. The lin- DISCUSSION ear regression equation slope for bone marrow cells stained A simple, reliable, and reproducible assay for HPC viability with AO/PI was –0.0063 over 2 hours, which indicates an is essential for clinical cellular engineering laboratories. extremely stable assay. TB-measured viability was less Traditionally, the TB dye exclusion assay has been used to stable over time. The stability of the AO/PI assay permits determine cell viability before and after cryopreservation staining and scoring of multiple specimens in batches and and before cell transfusion, but the TB assay can be diffi- also is helpful in training staff. cult to read and interpret. Moreover, despite widespread The AO/PI viability assay is a simple, rapid, highly lin- use, validation studies of the TB assay have not been pub- ear, functionally correlated assay that is superior to the con- lished. The AO/PI assay, with its two fluorescent dyes ventional TB viability assay. The AO/PI assay has not, how- viewed under darkfield fluorescence microscopy, permits ever, been validated with other HPCs, such as those derived clear differentiation of viable and nonviable cells. from peripheral or umbilical cord blood. The AO/PI assay has a higher degree of linearity and its results more closely resemble predicted viability than REFERENCES those of the TB assay. The TB assay was particularly prob- 1. Reeb BA. Dye exclusion test for bone marrow viability. In: lematic at lower percentages of cell viability. At 5-percent Areman EM, Deeg HJ, Sacher RA, eds. Bone marrow and and 25-percent predicted viability, the TB viability measure- stem cell processing: a manual of current techniques. New ments were 29.9 percent and 55.3 percent, respectively. Al- York: F.A. Davis Company, 1992:403-4. though most clinical specimens contain a higher percent- 2. Bank HL. Assessment of islet cell viability using fluorescent age of viable cells, some extensively manipulated specimens dyes. Diabetologia 1987;30:812-6. can fall within this range. This overestimation of viability 3. Jahanmehr SA, Hyde K, Geary CG, et al. Simple technique measurements by the TB assay can have clinical implica- for fluorescence staining of blood cells with acridine or- tions. ange. J Clin Pathol 1987;40:926-9. The percentage of viable cells measured with the AO/ 4. Bank HL. Rapid assessment of islet viability with acridine orange PI assay showed better correlation with CFU–GM frequency and propidium iodide. In Vitro Cell Dev Biol 1988;24:266-73. than did that measured by TB (r2 = 0.979 and r2 = 0.930, respec- 5. Thomas ED, Storb R. Technique for human marrow graft- tively). The TB assay overestimated progenitor cell function as ing. Blood 1970;36:507-15. measured by CFU–GM concentration. Clinical laboratories 6. Burger SR, Kadidlo D, McCullough J. Improved progenitor typically test specimens for cell viability more often than for assay standardization using peripheral blood progenitor CFU–GM frequency; it is helpful, therefore, to use a viabil- cells from a donor treated with granulocyte–colony-stimu- ity assay that correlates well with HPC assay results. lating factor. Transfusion 1999;39:451-6. 

696 TRANSFUSION Volume 40, June 2000