A rapid, semi-automated method for scoring micronuclei in mononucleated mouse lymphoma cells

A rapid, semi-automated method for scoring micronuclei in mononucleated mouse lymphoma cells

Mutation Research 726 (2011) 36–41 Contents lists available at SciVerse ScienceDirect Mutation Research/Genetic Toxicology and Environmental Mutagen...

790KB Sizes 0 Downloads 17 Views

Mutation Research 726 (2011) 36–41

Contents lists available at SciVerse ScienceDirect

Mutation Research/Genetic Toxicology and Environmental Mutagenesis journal homepage: www.elsevier.com/locate/gentox Community address: www.elsevier.com/locate/mutres

A rapid, semi-automated method for scoring micronuclei in mononucleated mouse lymphoma cells Ann T. Doherty a,∗ , Julie Hayes a , Mick Fellows a , Sarah Kirk b , Mike O’Donovan a a b

Genetic Toxicology, Safety Assessment, AstraZeneca R&D, Alderley Park, Macclesfield, Cheshire SK10 4TG, UK Statistics Group DECs Alderley Park, Macclesfield, Cheshire SK10 4TG, UK

a r t i c l e

i n f o

Article history: Received 2 March 2010 Received in revised form 14 June 2011 Accepted 30 July 2011 Available online 26 August 2011 Keywords: In-vitro micronuclei Semi-automated MicroNuc Population doubling

a b s t r a c t A semi-automated scoring system has been developed to provide rapid, accurate assessment of micronuclei in preparations of mononuclear mouse lymphoma L5178Y cells. Following exposure to a range of test agents, flat, single-cell preparations were produced from exponentially growing cultures by cytocentrifugation. Following staining with 4 -6-diamidino-2-phenylindole (DAPI), cells were scanned by use of the MicroNuc module of Metafer 4 v 3.4.102, after modifying the classifier developed for selecting micronuclei in binucleate cells to increase its sensitivity. The image gallery of all cells was then sorted to bring aberrant cells to the top of the gallery to assess visually the numbers of cells with micronuclei, as distinct from other debris. Slide quality was shown to be paramount in obtaining accurate results from an automated scan and the data obtained compared very well with the incidence of micronuclei scored conventionally by microscopy. Compared with manual scoring the time saving is considerable, as more than 2000 images are captured in approximately 2 min, with subsequent visual assessment of aberrant cells in the image gallery taking about 1–2 min/slide. By scanning all aberrant cells, the system also captures additional information on necrotic, apoptotic and fragmented cells. Although optimised for mouse lymphoma cells, it should be simple to adapt the method for any cell type growing in suspension. © 2011 Published by Elsevier B.V.

1. Introduction All cytogenetic analyses can be laborious and the assessment of micronuclei is used as a time-saving alternative to metaphase analysis. Furthermore, due to the uniform nature of micronuclei in cells at interphase, micronucleus scoring has the potential for automation to give further increased efficiency. Many attempts have been made to automate in vitro micronucleus scoring [1] and a number of systems are now commercially available for both slide scoring by use of image analysis, e.g., Metafer [2,3], Cellomics [4], and flow cytometry [5]. Although flow cytometry allows very large numbers of cells to be scored very quickly, it does not allow individual cells to be examined in detail to determine whether or not DNA fragments are really micronuclei and once scored, the samples cannot be reanalysed. It was decided to use an automated slide-scoring system in this laboratory (a) because visual inspection of the images allows micronuclei to be distinguished from other DNA fragments and artefacts and (b) semi-permanent preparations allow slides to be re-examined by conventional microscopy to clarify equivocal results. An automated system has been developed for monolayer cultures such as V79 and CHO cells [3] but this cannot be used for

∗ Corresponding author. Tel.: +44 1625 231285; fax: +44 1625 231281. E-mail address: [email protected] (A.T. Doherty). 1383-5718/$ – see front matter © 2011 Published by Elsevier B.V. doi:10.1016/j.mrgentox.2011.08.002

suspension cultures. The decision to use mouse lymphoma L5178Y cells for the in vitro micronucleus test in this laboratory was made in order to allow direct comparison with the results in the Tk mutation assay to be made when appropriate. The assessment of micronuclei in vitro has been established as an accepted method for determining genotoxicity and, in the regulatory setting, it is included as an alternative to either metaphase analysis or the mouse lymphoma Tk assay in the Step-2 revision of the ICH S2 (R1) guideline [6]. Both the draft OECD guideline 487 [7] and ICH S2 proposals allow suitably validated automated scoring systems to be used. Cytochalasin B is commonly used in the in vitro micronucleus assay in order to block cytokinesis and to identify micronuclei in binucleated cells, i.e., the population of cells that has divided [8–10]. This is particularly useful in cell lines with slow division rates and those – like primary cultures of lymphocytes – where only a proportion of the cells may be dividing; in rapidly dividing cultures, such as mouse lymphoma cells, the cytokinesis block may be unnecessary. Therefore, this work focused on a method to score micronuclei in mononucleate mouse lymphoma cells. The automated system used is the MicroNuc program on the Metafer platform by MetaSystems [2,3]. This program was initially written to detect micronuclei in binucleate cells [2,3], but contains classifiers that can be easily modified to determine the size and shape of nuclei and micronuclei and, therefore, could be

A.T. Doherty et al. / Mutation Research 726 (2011) 36–41

readily adapted to screen mononuclear cells. In order to compare the semi-automated system with conventional microscopy, mouse lymphoma cells were exposed to test agents with known genotoxic and/or cytotoxic mechanisms: mitomycin C (a cross-linking agent), triton X (a detergent), potassium bromate and paraquat (inducers of reactive oxygen), etoposide (a topoisomerase II poison), colchicine (an aneugen) and gamma radiation. 2. Materials and methods All chemicals and reagents were purchased from Sigma Aldrich (Dorset, UK) unless otherwise stated. 2.1. Test agents Mitomycin C (MMC), triton X, potassium bromate (KBrO3 ), paraquat and etoposide were all dissolved in dimethyl sulphoxide (DMSO); colchicine was dissolved in water. Gamma radiation was supplied from a caesium-137 source at a dose rate of 4.54 Gy/min. 2.2. Cell line and culture conditions Mouse lymphoma L5178Y Tk+/− cells, clone 3.7.2c, were obtained from Dr. J. Cole (MRC Cell Mutation Unit, University of Sussex, Brighton, UK). They were confirmed to be free from mycoplasma infection and had the expected karyotype, including two copies of chromosome 11 detected by FISH (in-house data). Cells were cultured in RPMI 1640 medium (Invitrogen, Paisley, UK) supplemented with 10% heat-inactivated donor horse serum (DHS), 2 mmol/L l-glutamine, 2 mmol/L sodium pyruvate, 1% Pluronic F68, 200 IU/mL penicillin, 200 ␮g/mL streptomycin and maintained at 37 ◦ C in a humidified atmosphere of 5% CO2 in air. 2.3. Treatment with test agents All compounds were tested in the absence of any exogenous metabolising system. For exposure to mitomycin C, triton X, potassium bromate, paraquat and etoposide, 107 cells were suspended in 20 mL RPMI medium containing 2.5% horse serum, and incubated for 3 h. For treatment with colchicine, 4 × 106 cells were suspended in 20 mL RPMI with 10% DHS and incubated for 24 h. All solvent and test solutions were added to the medium at 1% (v/v). For irradiation with ␥-rays, 107 cells were suspended in 20 mL RPMI containing 2.5% horse serum, and exposed for 20–106 s, corresponding with doses of 1.5–8 Gy. Duplicate cultures were prepared for each test-agent concentration and for the solvent control. Following treatment, the cells were centrifuged (200 × g), washed once and resuspended in 50 mL RPMI containing 10% DHS (for the 3-h exposure) or counted and adjusted to 2 × 105 /mL (for the 24-h exposure). Cultures were incubated for a further 24 h, then the cells were counted to estimate cytotoxicity, and microscope slides were prepared. 2.4. Slide preparation Cells (1 × 105 /mL) were suspended in medium containing 2% (v/v) Pluronic F68, and 850 ␮L of the suspension was dispensed into a Shandon Megafunnel® placed in the cytocentrifuge (Shandon Cytospin 3). The cells were centrifuged at 1000 rpm [give proper g-value here, Ed] for 8 min, then air-dried prior to fixation in 90% methanol for 10 min. Slides were mounted in Vectashield antifading agent containing 4 -6-diamidino2-phenylindole (DAPI) stain [10] with a large glass coverslip and stored flat in card trays, protected from light prior to analysis. 2.5. Cytotoxicity Cytotoxicity was estimated from the increase in cell number from the start of the treatment period until slide preparation, by use of the relative population doubling (RPD), which was determined as: RPD =

Number of population doublings in treated cultures × 100 Number of population doublings in control cultures

where population doubling = [log (cell number 24 h after the end of treatment/initial cell number)]/log 2. 2.6. Scanning Slides were scanned at 20× magnification with the Metafer 4 master station, comprising a Zeiss Axioplan Imager Z1, equipped with a Maerzhaeuser stepping motor stage that scans eight slides unattended. The MicroNuc module was run on the Metafer MSearch platform v3.4.102 (MetaSystems GmbH, Altlussheim Germany). Images were acquired on a peltier-cooled greyscale digital CCD camera Axiocam MRm (Carl Zeiss). The plane of focus was determined at a number of grid positions distributed evenly across the scan area. A predetermined scan area for the Shandon

37

Megafunnel was used for all slides and was set to avoid the outside margins of the cell-preparation area, as this area contains clumps of cells. The classifier was developed by MetaSystems to score binucleate cells, but it was modified to score mononucleates and altered in order to detect all aberrant cells for subsequent visual inspection. 2.6.1. MicroNuc settings The nucleus classifier for scanning mononuclear preparations was set to the following criteria: object threshold, 20%; minimum area, 10 ␮m2 ; maximum area, 400 ␮m2 ; maximum relative concavity of depth, 0.9; aspect ratio, 2.5; maximum distance between nuclei, 0 (as this feature is designed for scoring binucleates); maximum area asymmetry, 90%; region of interest radius, 40 ␮m; maximum object area in region of interest, 90 ␮m2 . The criteria for the micronuclei were set to: object threshold, 10%; minimum area, 1 ␮m2 ; maximum area, 55 ␮m2 ; maximum relative concavity of depth, 1; aspect ratio, 3.5; maximum distance, 35 ␮m. 2.7. Scoring A total of 2500 cells were captured per slide. All cells captured could be viewed in the image gallery. The image gallery was ordered so that all aberrant cells were visible at the top of the gallery. The assessment of micronuclei was made from the aberrant cells in the gallery taking between 1 and 2 min/slide. An assessment of error rate was conducted as some of the aberrant cells are not mononuclear and, therefore, may affect the micronucleus frequency: the error rate was found to be between 0.13% and 1.03% (control and positive control, respectively) and therefore has been dismissed. In parallel, 1000 cells/slide, from the same slides and from the same cultures, were analysed by conventional fluorescence microscopy. 2.8. Criteria for evaluation For both visual examination of Metafer images and microscopy, micronuclei were defined as having less than one third of the diameter of the main nucleus; being separate from the main nucleus with an intact membrane, and located within the cytoplasm area. Only cells that contained three or fewer micronuclei were counted. 2.9. Reproducibility of scoring An estimate of reproducibility was undertaken to assess variability between scorers using conventional microscopy or the Metafer system. For this purpose, five experienced cytogeneticists analysed the same eight slides (four solvent control DMSO and four positive control treated with 0.1 mmol/L 4NQO), both manually and with the MicroNuc method; this gives a total of 20 micronucleus values per slide, i.e., 10 from manual counts and 10 from MicroNuc. A scatter plot (Figs. 9 and 10) depicts the data used for the assessment of scoring variability. 2.10. Statistics for reproducibility of scoring In the comparison between scoring methods, the positive control and negative control data were modelled separately due to the differences in variability between the two controls. As the manual counts were performed on 1000 cells while the automated counts were done on a variable number of cells, ranging from 2249 to 2306, the automated counts were first converted to a micronuclei count per 1000 cells prior to further analysis. In addition, the counts (per 1000 cells) were transformed by use of an average square-root transformation, since this is the scale on which all statistical analyses are routinely performed. A mixed model was fitted in SAS (a commercially available statistical package) with PROC MIXED, where method (manual or automated), slides (1–4) and readers (1–5) were fitted as categorical variables. Method was fitted as a fixed effect and slide, reader, all two-way interactions between slide, reader and method and the three-way interaction, method-slide-reader, were all fitted as random effects. Four replicate readings, taken on different occasions, for each slide/reader/method combination were used to estimate error. This analysis was used to generate the mean count level associated with the manual and automated methods, and the magnitude of the different components of variation. Separate estimates of variation associated with the manual and the automated methods were then generated in SAS by use of PROC VARCOMP, and fitting slide, reader, and the interaction slide-reader as random effects. Finally, the Z-factor (or Z-prime) statistics was calculated as 1 − S/D for both the manual and automated methods. For each method, D is the difference between the average transformed counts for the positive and negative controls. When routine in vitro micronucleus studies are run, two results are generated for each type of control, positive or negative, and the duplicate mean is calculated. From our variation analysis in this study, we can estimate the standard error associated with routine control duplicate means. For each method, S is three times the sum of the positive and negative control standard errors. The Z-prime or Z-factor is used to evaluate high-throughput screening assays. By definition, its value is less than one, with values closer to one being preferred, and achieved as the difference between means increases, or the variation decreases. A value of 0.5 or greater is usually required for an assay to be considered good.

38

A.T. Doherty et al. / Mutation Research 726 (2011) 36–41

Figs. 1–7. Bar graphs with semi-automated micronucleus (semi-automated) score alongside conventional microscopy (microscopy), presented as percentage of micronuclei for each individual test agent. The line graph included indicates the relative population doubling (RPD) as proportion of the individual control values presented as percent survival.

A.T. Doherty et al. / Mutation Research 726 (2011) 36–41

39

Fig. 8. Images from the image gallery with examples of a micronucleated cell, an apoptotic cells, a necrotic cells and an example of fragmentation.

3. Results The success of the automated scanning system was found to be dependent on a number of technical steps in order to generate highquality slide preparations, most importantly the use of serum-free RPMI with 2% pluronic F68 to prevent cells from clumping. Cell samples must also be freshly prepared and immediately dispensed into cytocentrifuge funnels to keep them as single-cell suspension. The use of Megafunnels rather than the standard 200 ␮L funnels results in better single-cell distribution over a larger area, thus increasing the number of cells that can be scored. The data presented in Figs. 1–7 show very good correlation between the Metafer scores and conventional microscopical analysis of the same slides. It is important to note that the Metafer scan was configured to detect all aberrant cells, but the data in Figs. 1–7 show only the micronucleus counts after visual inspection of the galleries. The results for MMC, etoposide, KBrO3 and ␥-radiation show statistically significant, concentration-related increases in micronuclei by both methods (Figs. 1–4). In general, the semi-automated counts were lower than those obtained by microscopical examination, probably because some micronuclei very close to the main nuclei cannot be detected by the automated system. However, the threshold concentrations at which significant increases in micronucleated cells were seen were the same for both methods. Colchicine (Fig. 5) induced statistically significant increases in the number of micronucleated cells at the highest concentrations analysed in the semi-automated analyses, but not in microscopical analysis. This effect appears to be unique to colchicine of the agents examined here and may be due to the high numbers of fragmented cells observed. Triton X (Fig. 6) and paraquat (Fig. 7) were included as cytotoxic, non-genotoxic agents and neither showed a significant increase in micronucleus frequency at any concentration analysed by either method. The automated scan detected cell debris in aberrant cell divisions (Fig. 8) and the increases were directly related to toxicity as measured by RPD. Fig. 8 shows examples of the image gallery after the slides have been scanned: a cell with two micronuclei, an apoptotic cell, a necrotic cell, and a fragmented cell. Comparisons of both scoring methods are shown in the scatter plot (Figs. 9 & 10). For both negative and positive controls, the largest source of variation was residual (i.e., not associated with specific sources such as “slide”), with the interaction between reader and method producing the next greatest contribution, followed by the interaction between slide and method. The consistency in the pattern across positive and negative controls, but the indication that the variation associated with different readers and slides was not consistent across methods confirmed that a separate manual/automated analysis of variance was appropriate to obtain

estimates for use in calculating the Z-factor (Z-prime) statistics (Table 1). The Z-prime value 0.50 is, in fact, an underestimate for the manual method, which usually involves counting 2000 cells, with less-variable results than for the 1000 cells counted in this study (Table 2). The Z-prime value, based on two slides per group, is acceptable for both the automated and the manual count method, i.e., both have an acceptable window, relative to variation, between the positive and negative (vehicle) controls (Table 3). What this analysis cannot do is assess how the bias between manual and automated methods changes across count levels, and how this impacts the measured value of the biologically relevant effect size between vehicle and treated groups that we wish to be able to detect routinely. However, the in vitro dose–response data illustrated elsewhere in this paper indicates that the bias is worst at the highest levels, for known biological reasons. Provided the automated method is only used as a dose–response screen for clear positives, the automated-count bias seems unlikely to cause problems in practice. The data indicate that the variability between individual scorers is similar using either method. However, as may be expected, the reproducibility of repeated scoring of the same slide is increased with the automated platform. Table 1 Pattern of variation in results. From the mixed-model analysis of variance, the largest contributions to the overall variability in the data. Negative control (vehicle) Var (error) Var (reader × method) Var (slide × method)

Positive control (4 NQO) 64% 24% 7%

Var (error) Var (reader × method) Var (slide × method)

53% 28% 13%

Table 2 Assessment of assay window for manual counts. From the models, the following estimates were obtained for the manual counts. Control

Mean

Standard deviation

Standard error of the mean (n = 2)

Negative Positive

1.398 8.180

0.544 1.044

0.384 0.738

D, difference between means = 6.782; sum of std. errors = 1.122; S = 3 × sum of std. errors = 3.366; Z-prime (1 − S/D) = 0.50.

Table 3 Assessment of assay window for automated counts. From the models, the following estimates were obtained for the automated counts. Control

Mean

Standard deviation

Standard error of the mean (n = 2)

Negative Positive

1.280 5.550

0.325 0.433

0.230 0.306

D, difference between means = 4.270; sum of std. errors = 0.536; S = 3 × sum of std. errors = 1.608; Z-prime (1 − S/D) = 0.62.

40

A.T. Doherty et al. / Mutation Research 726 (2011) 36–41

Figs. 9 and 10. Scatter plots in which each individual scorer is represented on the x-axis as nos. 1–5 (reader) against percent micronuclei (% MN). Individual counts by conventional microscopy are shown as black circles (䊉) and semi-automated counts as a grey triangles ( ). Four individual slides were examined from both vehicle control (DMSO) (Fig. 8) and positive control (0.1 mmol/L 4NQO) (Fig. 9).

A.T. Doherty et al. / Mutation Research 726 (2011) 36–41

4. Discussion The Metafer image-analysis system was able to identify damaged mouse lymphoma cells and manual assessment of the captured images enabled micronuclei to be quickly distinguished from other debris. Good concordance was obtained with the results obtained by conventional microscopical scoring of the same slides for all four agents investigated. Considerable time-saving was achieved with more than 2000 images captured in approximately 2 min and visual assessment of the image gallery taking a further 1–2 min/slide. Furthermore, some of the inter-individual variation seen between scorers was reduced. Finally, because of the saving in time, the statistical power can be increased by routinely scoring 2500 or more cells rather than the 1000 cells/slide by microscopy and the number can be increased further to elucidate equivocal or marginal responses. In addition the increased numbers of cells scored may reduce some of the inter-individual variation noted when assessing chromosomal damage [11]. A possible limitation of this system is that it detects slightly fewer micronuclei than does conventional scoring, probably due to the intensity of the staining of the main nuclei and the small size of micronuclei, or because some micronuclei overlap each other. However, these minor differences are consistent and did not alter, for any of the genotoxins tested, the minimum concentrations giving significant increases in micronucleus frequency. Although Metafer can be used as a fully automated platform, it was decided in this laboratory to use it as a tool to find potentially micronucleated cells, with the final assessment being performed by experienced cytogeneticists. A semi-automated approach has also been preferred for scoring micronuclei in binucleate lymphocytes [12]. A significant advantage of image analysis over flow cytometric methods for scoring micronuclei is that additional information is provided on cytotoxic cellular damage such as apoptosis, necrosis and fragmented cells and, unlike flow cytometry, micronuclei scored can be distinguished from cellular debris. Also, the ability to detect necrotic and apoptotic bodies may provide additional information on the mechanism of action of a test compound [8]. A further advantage is that this system can be used to find micronuclei when a centromere-specific label has been applied to investigate aneugenicity and, in this respect, the ability to find rapidly all micronucleated cells is particularly valuable. Although it has not been necessary for a micronucleus test with mouse lymphoma cells, it should be possible to modify the MicroNuc program to score both mononucleate and binucleate cells by performing two scans (i.e., first for mononucleate, then for binucleate cells) or by adding a second stain for the cell membrane. In conclusion, the Metafer semi-automated system provides a rapid, accurate method to score micronuclei in mouse lymphoma cells. In principle, the method should be simple to adapt for any cells growing in suspension and work is in progress with TK6 cells

41

and primary human lymphocytes. Although it is not intended to use the semi-automated method as a high-throughput screen in this laboratory, in support of early drug-discovery projects it has a significant advantage over the Tk-mutation assay in that it requires much smaller quantities of test compound – i.e., ∼5 mg compared with ∼200 mg – to achieve concentrations up to 1 mmol/L. Conflict of interest None. Funding Astrazeneca Pharmaceuticals only. Acknowledgement The authors would like to thank Sean Evans, Joe Ryan and Katie Clare for technical assistance. References [1] W. Böcker, C. Streffer, W.-U. Müller, C. Yu, Automated scoring of micronuclei in binucleated human lymphocytes, Int. J. Radiat. Biol. 70 (1996) 529–537. [2] D. Varga, T. Johannes, S. Jainta, S. Suchuster, U. Schwarz-Boeger, M. Kiechle, B.P. Garcia, W. Vogel, An automated scoring procedure for the micronucleus test, Mutagenesis 19 (2004) 391–397. [3] C. Schunck, T. Johannes, D. Varga, T. Lörch, A. Plesch, New developments in automated cytogenetic imaging: unattended scoring of dicentric chromosomes, micronuclei, single cell gel electrophoresis, and fluorescence signals, Cytogenet. Genome Res. 104 (2004) 383–389. [4] D. Diaz, A. Scott, P. Carmichael, W. Shi, C. Costales, Evaluation of an automated in vitro micronucleus assay in CHO-K1 cells, Mutat. Res. 630 (2007) 1–13. [5] S.M. Bryce, J.C. Bemis, S.L. Avlasevich, S. Dertinger, In vitro micronucleus assay scored by flow cytometry provides a comprehensive evaluation of cytogenetic damage and cytotoxicity, Mutat. Res. 630 (2007) 78–91. [6] International Committee on Harmonisation ICH S2(R1): Guidance on Genotoxicity Testing and Data Interpretation for Pharmaceuticals Intended for Human Use. http://www.ich.org/cache/compo/502-272-1.html#S2. Step 4 of revision process (accessed 20.12.09). [7] Organisation for Economic Co-Operation and Development, OECD Guideline for Testing Chemicals 487: In Vitro Micronucleus Test. http://www.oecd.org/ dataoecd/33/22/37865944.pdf, approved 21 July 2010. [8] M. Fenech, J. Crott, J. Turner, S. Brown, Necrosis, apoptosis, cytostasis and DNA damage in human lymphocytes measured simultaneously within the cytokenisis-block micronucleus assay: description of the method and result for hydrogen peroxide, Mutagenesis 14 (1999) 605–612. [9] S. Albertini, B. Miller, A.-A. Chételat, F. Locher, Detailed data on in vitro MNT and in vitro CA: industrial experience, Mutat. Res. 392 (1997) 187–208. [10] Ph. Castelain, P. Van Hummelen, A. Deleener, M. Kirsch-Volders, Automated detection of cytochalasin-B blocked binucleated lymphocytes for scoring micronuclei, Mutagenesis 8 (1993) 285–293. [11] R. Mateuca, N. Lombaert, P.V. Aka, I. Decordier, M. Kirsch-Volders, Chromosomal changes: induction, deletion, detection methods and applicability in human biomonitoring, Biochime 88 (2006) 1515–1531. [12] H. Thierens, A. Vral, F. De Scheerder, L. De Ridder, A. Tates, Semi-automated micronucleus scoring in cytokinesis-blocked lymphocytes and after irradiation, Int. J. Radiat. Biol. 72 (1997) 319–324.