Interlaboratory assessment of mitotic index by flow cytometry confirms superior reproducibility relative to microscopic scoring
Abstract
A flow cytometric procedure for determining mitotic index (MI) as part of the metaphase chromosome aberrations assay, developed and utilized routinely at Pfizer as part of their standard assay design, has been adopted successfully by Covance laboratories. This method, using antibodies against phosphorylated histone tails (H3PS10) and nucleic acid stain, has been evaluated by the two independent test sites and compared to manual scoring. Primary human lymphocytes were treated with cyclophosphamide, mitomycin C, benzo(a)pyrene, and etoposide at concentrations inducing dose-dependent cytotoxicity. Deming regression analysis indicates that the results generated via flow cytometry (FCM) were more consistent between sites than those generated via microscopy. Further analysis using the Bland–Altman modification of the Tukey mean difference method supports this finding, as the standard deviations (SDs) of differences in MI generated by FCM were less than half of those generated manually. Decreases in scoring variability owing to the objective nature of FCM, and the greater number of cells analyzed, make FCM a superior method for MI determination. In addition, the FCM method has proven to be transferable and easily integrated into standard genetic toxicology laboratory operations. Environ. Mol. Mutagen. 2012. © 2012 Wiley Periodicals, Inc.
INTRODUCTION
The in vitro chromosome aberration assay can be used as part of the current genotoxicity test battery in pharmaceutical development (ICH,2011). In the absence of solubility limitations, guidelines require testing to the limit dose (1 mM or 0.5 mg/mL), or to a concentration inducing approximately 50% reduction in cell viability (OECD,1997). When using primary human lymphocyte cultures, the de facto method for this cytotoxicity determination is mitotic index (MI). Historically, this has required manual microscopic scoring to determine the percentage of cells in mitosis. However, an automated flow cytometric method has been developed, which utilizes antibodies against histone H3 proteins phosphorylated at serine residue 10 (H3PS10; Muehlbauer and Schuler,2003).
H3PS10 is globally expressed by eukaryotes during meiosis and mitosis, and is also present to a lesser extent during apoptosis and transcription (Prigent and Dimitrov,2003). Although phosphorylation is a requirement for chromosome condensation in mitosis, apoptotic chromosome condensation occurs in the presence of unphosphorylated H3 proteins, which has been attributed to increased protein kinase A activation (Yoshida et al.,1997). Similarly, general transcription does not induce global phosphorylation of H3S10, and meiotic H3PS10 events are not relevant to the primary human peripheral blood lymphocytes used in this cross-site evaluation.
Prior to engaging in this project, both laboratories independently validated the use of flow cytometry (FCM) to assess MI by performing direct intralaboratory comparisons against manual microscopy (Muehlbauer and Schuler,2003; Roberts et al.,2008). Correlation between the two methodologies was good, but manual microscopy suffers from inherent subjectivity, scorer-to-scorer variability, and variations in slide preparation. MIs of cultures determined by FCM were highly concordant, whereas the results obtained manually lacked consistency between cultures treated with the same dose, and were dependent on the area of the slide scored and the person performing the evaluation (Muehlbauer and Schuler,2003). The objective of this interlaboratory evaluation was to evaluate the consistency of FCM data generated between two sites, and to assess the transferability of the FCM method.
MATERIALS AND METHODS
The culturing, labeling, and analysis procedures used herein were standardized across laboratories.
Cell Culture
Human venous blood from healthy adult donors (nonsmokers, aged 18–40, without a history of radiotherapy, chemotherapy, or drug usage, and lacking current viral infections) was drawn into sterile, heparinized Vacutainers® (BD, Franklin Lakes, NJ). Whole blood was diluted with an equal volume of PBS (Mediatech, Manassas, VA), and mononuclear cells (MNCs) were isolated via centrifugation using Lymphoprep® tubes (Axis-Shield, Dundee, Scotland). MNCs were washed with PBS and cultured in Williams medium E (WME; Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (Quality Biologicals, Gaithersburg, MD), 2 mM L-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin (all from Mediatech). Cultures were established in upright 75-cm2 flasks, stimulated with phytohemagglutinin-M (1%; Invitrogen), and incubated at 37°C in a humidified atmosphere of 5% CO2 in air for approximately 48 hr with intermittent shaking.
Prior to treatment, cultures were pooled by donor and centrifuged. The supernatant was aspirated and the cells were resuspended in serum-free WME (WME−). A cell count was performed and additional WME or WME− was added as necessary to prepare cultures to be treated in the presence or absence of metabolic activation (S9), respectively. Final cell density was 2.0–3.5 × 105 cells/mL in a final volume of 5 mL.
Chemicals
Dimethyl sulfoxide (DMSO; CAS 67-68-5), benzo(a)pyrene (B(a)P; CAS 50-32-8), cyclophosphamide (CP; CAS 6055-19-2), mitomycin C (MMC; CAS 50-07-7), and etoposide (ETD; CAS 33419-42-0) were obtained from Sigma-Aldrich (St. Louis, MO). All test chemicals were prepared in DMSO at 100-fold the final desired treatment concentrations and administered using a 1% dose volume.
Selection of MMC and CP was based on their historical and recommended use as cross-linking positive control agents in the chromosome aberration assay (OECD,1997). ETD and B(a)P were chosen because of their alternate mechanisms of action (topoisomerase II inhibition and covalent DNA intercalation, respectively (King et al.,1975; Sieber et al.,1976; Xu et al.,2009). Dose selection for this study was based on unpublished historical data and previously reported results (Muehlbauer and Schuler,2005). The highest concentrations evaluated were expected to induce ≥50% relative decreases in MI, as compared to respective vehicle control-treated cultures.
Treatment
CP and MMC treatments were performed at Covance, whereas B(a)P and ETD treatments were performed at Pfizer. CP and B(a)P treatments were performed in the presence of an exogenous metabolic activation system containing Aroclor 1254-induced rat liver homogenate (S9 fraction) and 0.5 mM NADPH (Moltox, Boone, NC). The final concentrations of liver homogenate in the CP and B(a)P treatments were 0.90 and 0.27% (v/v), respectively. The difference in S9 concentration between test sites was a consequence of lot-specific titration, demonstrated by the biotransformation of CP (10 μg/mL) into an active mutagenic metabolite producing a biologically significant increase in chromosomal aberrations. All treatments were performed in duplicate cultures.
Three-hour treatments were performed for all chemicals tested with or without S9 as indicated; continuous 24-hr treatments were also performed for ETD and MMC (Table I). After the 3-hr treatments, cultures were centrifuged, washed, refed with fresh WME, and reincubated until harvest. All cultures were harvested 24 hr after the start of treatment, with Colcemid® (final, 0.075 μg/mL,; Invitrogen) added approximately 3 hr prior to harvest.
| Test site | Chemical | Maximum concentration (μg/mL) | Treatment timea (hr) | Liver homogenate (%, v/v; final) |
|---|---|---|---|---|
| Covance | CP | 15.0 | 3 | 0.90 |
| MMC | 0.600 | 3 | — | |
| 0.125 | 24 | — | ||
| Pfizer | B(a)P | 100 | 3 | 0.23 |
| ETD | 10.0 | 3 | — | |
| 0.500 | 24 | — |
- a Colcemid® was added at 21 hr and all cultures were harvested at 24 hr.
FCM Harvest, Labeling, and Staining
Duplicate 1–2 mL samples of each culture were transferred to 5-mL round bottom polystyrene tubes (BD, Franklin Lakes, NJ) containing 2 mL cold (4°C) PBS. After centrifugation, cell pellets were resuspended in 200 μL cold PBS and the cells were fixed/stored at least overnight in 2–4 mL ice-cold (−20°C) 70% (v/v) ethanol. One replicate FCM sample was shipped overnight, upright, in an insulated container filled with wet ice and ≤−20°C ice packs to the other test site, whereas the remaining samples were stored at −20 ± 10°C. FCM samples were centrifuged, the supernatants were decanted, and the cell pellets were resuspended. Samples were rehydrated and permeabilized with 0.01% Tween-80 (Sigma-Aldrich) in Ca/Mg- and phenol red-free PBS (Mediatech). Mitotic cells were labeled with mouse anti-H3PS10 (diluted 1:400 in PBS) monoclonal antibodies (Cell Signaling Technologies, Beverly, MA) for 45–60 min at 37°C. Samples were washed with PBS and stained with Alexa Fluor® 488-conjugated goat anti-mouse IgG (AF488; Invitrogen) to label the anti-H3PS10 antibodies (diluted 1:1,000 in PBS). Samples were then washed, centrifuged, and the nucleic acids were stained with 25 μg/mL propidium iodide (PI) in the presence of 0.30 mg/mL Ribonuclease A (both from Sigma-Aldrich).
Cytogenetic Harvest, Slide Preparation, and Scoring
The remaining culture volumes were transferred to 15-mL tubes and centrifuged. These pellets were resuspended and the cells were swelled in 4–6 mL of 75 mM potassium chloride (Sigma-Aldrich), and prefixed in 2 mL methanol:glacial acetic acid (3:1 v/v; Malinckrodt Baker, Phillipsburg, NJ). After centrifugation, 5 mL of fresh fixative was added and the cells were stored at 4°C. Slides (at least in duplicate) were prepared by centrifuging and resuspending the cells in <1 mL of supernatant, and dropping the cell suspensions onto glass slides. Slides were stained with 5% Giemsa (Richard-Allen Scientific, Kalamazoo, MI), air dried, and permanently mounted. MI was determined by two independent scorers at Covance, each counting 1,000 cells/slide per duplicate culture.
FCM Hardware and Software
Covance used the E and D channels of a FACSCanto II to collect AF488 and PI fluorescence, respectively (digital cytometer, FACSDiva software, v6.1.1; E channel: 502–560 nm; D channel: 556–627 nm). Pfizer used the FL1 and FL2 channels of a FACSCalibur to collect AF488 and PI fluorescence, respectively (analog cytometer, CellQuest software, v3.3; FL1 channel: 515–545 nm; FL2 channel: 564–606 nm). In both cases, excitation was with a 488-nm argon laser, and all hardware and software was acquired from BD Biosciences (San Diego, CA). Both cytometers had original factory configurations.
FCM Data Acquisition
The flow cytometers were calibrated on each day of analysis, and a harmonized experimental template was used to acquire data, as described previously (Muehlbauer and Schuler,2003,2005). When possible, the samples were acquired using a low-speed setting (<500 events/sec), and a FSC threshold was employed. The PI-channel voltage was set accordingly to align cells with G0/G1 DNA content within the predefined interval gate (Fig. 1, top). This gate was only used for alignment to ensure proper subsequent gating. Single-nucleated cells were gated using a dot plot of PI width versus area (Fig. 1, middle), and the H3PS10+ population then was identified using a bivariate dot plot of PI area against AF488 fluorescence (Fig. 1, bottom). A gate was placed around the H3PS10+ population and the % MI was calculated as the number of H3PS10+ events divided by the number of cycling cells × 100. Minor manual compensation was used to eliminate AF488 emissions from the detector collecting PI fluorescence. The stop gate for data acquisition was 2,000 mitotic events, if available.
Representative vehicle control culture. Top: DNA histogram used for alignment to ensure accuracy of subsequent gating (showing Cycling Cells, y-axis is cell count). Middle: Doublet discrimination using PI-W vs PI-A. Gate widens to accommodate cells with > 4N DNA content. Bottom: Mitotic Cells gate drawn for H3PS10 positive cells with > 2N G0/G1 DNA content. Plots are zoomed in for clarity, end of linear X-axis not shown. Cycling cells gate does not extend to the end of the plot, eliminating events that stack in the highest channel. Gating hierarchy from BD FACSDiva v6.1.1.
Data Analyses
Relative mitotic suppression (RMS) was calculated using mean MIs and the following equation: [1−(MItreated/MIcontrol)]×100. Data from replicate cultures were then subjected to unweighted Deming regression analysis (Linnet,1998), and the Bland–Altman modification of the Tukey mean difference method (Bland and Altman,1986).
RESULTS
Simple visual comparison of the data indicates that the RMS results obtained by FCM were virtually indistinguishable between sites in almost all cases (Figs. 2-4). The lone exception was for cultures treated with B(a)P, largely owing to a difference in the absolute MI value obtained for the vehicle controls (10.8 at Pfizer versus 15.2 obtained by Covance; Fig. 4b). In contrast, much greater variability was observed in RMS values obtained by microscopic scoring. These differences in manual scoring were most notable for the 3-hr treatments with MMC and B(a)P (Figs. 2a and 4b, respectively).
RMS determined by FCM and microscopy for cultures treated with Mitomycin C for 3- or 24-hours (A and B, respectively). Open symbols represent FCM data acquired at Covance (triangles) or Pfizer (circles). Solid symbols are microscopic data generated by two independent scorers. The dashed line indicates acceptable levels of cytotoxicity (i.e., 50% RMS).
RMS determined by FCM and microscopy for cultures treated with Etoposide for 3- or 24-hours (A and B, respectively). Symbols and lines are the same as described in Figure 2 legend.
RMS determined by FCM and microscopy for cultures treated with Cyclophosophamide or Benzo(a)pyrene for 3-hours in the presence of rat liver homogenate (A and B, respectively). Symbols and lines are the same as described in Figure 2 legend.
Deming regression was used to compare each methodology between test sites. This statistical approach is more appropriate for analytical comparisons than standard linear regression, as error in both variables is taken into account. With this method, an observed ratio of variances (Sy|x) equal to 1.00 indicates a perfect fit. A high correlation (Sy|x = 1.16) was found between FCM data generated by the two sites (corresponding r2 of 0.899; Fig. 5a), and the SD of differences in MI between sites was 1.17 (Fig. 5b). Data generated via microscopy had a lower correlation between scorers (Sy|x = 2.23, corresponding r2 = 0.684, Fig. 5c) and the SD of differences in MI was approximately twofold higher at 2.36 (Fig. 5d). These results indicate that FCM is much more consistent for measuring MI than microscopy.
Left column: Demming's regression analyses conducted on data generated across sites via FCM (A), microscopy (C), and pooled data (E). Solid lines indicates perfect correlation (Y=X), while Demming's regression is represented by the dashed line. The panels in the right column are Bland-Altman representations of the same data (panels B, D, and F represent FCM, microscopy, and pooled data, respectively). The mean difference in MI is represented by the bold dashed line with 95% confidence intervals marked as thin dashed lines. Solid line (Y = 0) indicates no difference between sites (B, D) or methodology (F). Open circles represent replicate culture data, SD = Standard Deviation, Sy∥x = ratio of variances (y/x).
FCM and microscopy were also compared to determine the correlation between both techniques. Data from all experiments were pooled based on method, and Deming regression indicates fairly good correlation (Sy|x = 1.87, corresponding r2 = 0.777, Fig. 5e). The SD of differences in MI was 1.81 (Fig. 5f), which is less variable than when comparing the two manual scorers against each other (Fig. 5d).
DISCUSSION
According to the current guidelines, (ICH,2011; OECD,1997), cultures should be scored for chromosomal aberrations only up to ∼ 50% RMS, as the evaluation of excessively toxic doses may lead to false-positive results (Galloway,2000). Inadvertent omission of condensed G0/G1 cells during manual scoring will consequently inflate the MI, potentially masking true cytotoxicity, as shown in Figure 5e in which MIs generated by microscopy were consistently higher than those generated by FCM. An additional hurdle to accurate manual scoring is differences in cell deposition on the slides, as scoring different parts of the same slide can yield different MIs (Muehlbauer and Schuler,2003).
The maximum concentration evaluated between sites for the 3-hr MMC treatment would have been identical if based on MIs generated by FCM (Fig. 2A, open symbols). In contrast, reliance upon the microscopic results generated by Scorer B from the 3-hr treatment would have yielded evaluation of a very different maximum concentration as suitable toxicity was not observed throughout the dose range evaluated (Fig. 2a, solid squares). This example clearly demonstrates the inherent scorer-to-scorer variability, which is a consequence of subjective observations, differences in slide preparation and staining characteristics, and can depend on the area of the slide evaluated. Similarly, differences in FCM and microscopic MIs for cultures treated with CP (Fig. 4a) would produce approximately twofold differences in the concentration ranges scored for chromosomal aberrations. Such differences could lead to irrelevant, false-positive results when testing compounds whose clastogenic profile is unknown.
The raw MI values obtained by FCM for B(a)P-;treated cultures were quite similar between sites, yet relatively large differences in RMS were observed owing to the difference in vehicle control MIs (Fig. 4b). These data demonstrate the need for accurate MI measurements, especially in vehicle controls, as they can have dramatic effects on selecting appropriate doses for chromosomal aberration analysis. Still, RMS determined by all four independent measurements reached a plateau by approximately 20 μg/mL B(a)P. Similar results were observed in a subsequent experiment using even higher concentrations of B(a)P (data not shown). Considering that MI is a measurement of proliferation and that B(a)P requires metabolic activation, this threshold-like response may reflect metabolic limitations owing to S9 concentration and/or treatment time, or the limited proportion of asynchronous cells passing through a critical cell-cycle stage during the treatment period. This example illustrates that although the FCM platform is overall more consistent in measuring MIs, it is not resilient to the consequence that differences in MIs of vehicle control-treated cultures have on calculating RMS.
Another drawback to the FCM platform is the cost of instrumentation, maintenance, and operator training. Overall, the time to generate MI data remains the same, with the major advantage being that less slides are prepared, and quantitatively the data are more consistent and objective, with replicate cultures typically having MIs within 1% of each other. Once cytotoxicity measurements are acquired on the flow cytometer, slides for chromosomal aberration analysis need only be prepared for the doses of interest (i.e., controls and test article doses approximating 50% cytotoxicity, as applicable). The automated nature of data acquisition with the FCM platform allows for more doses to be routinely analyzed, which facilitates a better evaluation of the dose response curve.
Upon adoption of this platform to score MI, investigators can further evaluate the DNA content of the H3PS10+ population to determine the percent of mitotic cells in aneuploid states (Muehlbauer and Schuler,2005). Although polyploidy and endoreduplication are commonly scored manually, this evaluation is made only by scoring a few hundred mitotic cells per dose level. Flow cytometric evaluation of MI allows thousands of cells to be objectively and quantitatively analyzed for aneuploidy (Muehlbauer et al.,2008), which inherently increases the statistical power of the endpoint. Additionally, qualitative cell-cycle data are easily evaluated using the current platform and can help determine test article induced cytostasis and perturbations to cell-cycle kinetics, both useful in elucidating the mechanism of genotoxicity. Furthermore, if desired, the cell-cycle histogram can be analyzed by commercially available software to quantitatively determine dose-related cell-cycle perturbations.
CONCLUSION
MI determined via FCM is less variable than traditional microscopic measurement. The FCM method has proven to be transferable between laboratories and across digital and analog flow cytometers, and can be easily integrated into standard genetic toxicology laboratory operations. The objective nature of FCM and the greater number of cells analyzed makes FCM a superior method for consistent MI measurements, which has the potential to minimize concerns regarding the accuracy of dose selection for the in vitro chromosome aberrations assay.
CONFLICT OF INTEREST
Covance Laboratories offers this assay as a contracted service.
