An imaging flow cytometric method for measuring cell division history and molecular symmetry during mitosis


  • Andrew Filby,

    Corresponding author
    1. Flow Cytometry Laboratory, London Research Institute, Cancer Research UK, London, WC2A 3LY, United Kingdom
    • Flow Cytometry Laboratory, London Research Institute, Cancer Research UK, London, WC2A 3LY, UK===

    Search for more papers by this author
  • Esperanza Perucha,

    1. Division of Transplantation, Immunology and Mucosal Biology and MRC Centre for Transplantation, Guy's and St Thomas Hospital, King's College, London, SE1 9RT, United Kingdom
    2. Immune Monitoring Laboratory, NIHR Biomedical Research Centre at Guy's and St Thomas' NHS Foundation Trust and King's College London, United Kingdom
    Search for more papers by this author
  • Huw Summers,

    1. Multidisciplinary Nanotechnology Centre, School of Engineering, Swansea University, Swansea, SA2 8PP, United Kingdom
    Search for more papers by this author
  • Paul Rees,

    1. Multidisciplinary Nanotechnology Centre, School of Engineering, Swansea University, Swansea, SA2 8PP, United Kingdom
    Search for more papers by this author
  • Prabhjoat Chana,

    1. Immune Monitoring Laboratory, NIHR Biomedical Research Centre at Guy's and St Thomas' NHS Foundation Trust and King's College London, United Kingdom
    Search for more papers by this author
  • Susanne Heck,

    1. Immune Monitoring Laboratory, NIHR Biomedical Research Centre at Guy's and St Thomas' NHS Foundation Trust and King's College London, United Kingdom
    Search for more papers by this author
  • Graham M. Lord,

    1. Division of Transplantation, Immunology and Mucosal Biology and MRC Centre for Transplantation, Guy's and St Thomas Hospital, King's College, London, SE1 9RT, United Kingdom
    2. Immune Monitoring Laboratory, NIHR Biomedical Research Centre at Guy's and St Thomas' NHS Foundation Trust and King's College London, United Kingdom
    Search for more papers by this author
  • Derek Davies

    1. Flow Cytometry Laboratory, London Research Institute, Cancer Research UK, London, WC2A 3LY, United Kingdom
    Search for more papers by this author


Asymmetric cell division is an important mechanism for generating cellular diversity, however, techniques for measuring the distribution of fate-regulating molecules during mitosis have been hampered by a lack of objectivity, quantitation, and statistical robustness. Here we describe a novel imaging flow cytometric approach that is able to report a cells proliferative history and cell cycle position using dye dilution, pH3, and PI staining to then measure the spatial distribution of fluorescent signals during mitosis using CCD-derived imagery. Using Jurkat cells, resolution of the fluorescently labeled populations was comparable to traditional PMT based cytometers thus eliminating the need to sort cells with specific division histories for microscopy. Subdividing mitotic stages by morphology allowed us to determine the time spent in each cell cycle phase using mathematical modeling approaches. Furthermore high sample throughput allowed us to collect statistically relevant numbers of cells without the need to use blocking agents that artificially enrich for mitotic events. The fluorescent imagery was used to measure PKCζ protein and EEA-1+ endosome distribution during different mitotic phases in Jurkat cells. While telophase cells represented the favorable population for measuring asymmetry, asynchronously dividing cells spent approximately 43 seconds in this stage, explaining why they were present at such low frequencies. This necessitated the acquisition of large cell numbers. Interestingly we found that PKCζ was inherited asymmetrically in 2.5% of all telophasic events whereas endosome inheritance was significantly more symmetrical. Furthermore, molecular polarity at early mitotic phases was a poor indicator of asymmetry during telophase highlighting that, though rare, telophasic events represented the best candidates for asymmetry studies. In summary, this technique combines the spatial information afforded by fluorescence microscopy with the statistical wealth and objectivity of traditional flow cytometry, overcoming the key limitations of existing approaches for studying asymmetry during mitosis. © 2011 International Society for Advancement of Cytometry

Determining how fate decisions are made at the molecular level is key to understanding how single cell progenitors can generate different specialized cell types. This diversification can occur in two ways. When a cell divides by mitosis, equal distribution of fate determining molecules across the plane of cytokinesis will produce two identical daughter cells that can differentiate in response to extrinsic stimuli. Alternatively, if fate determinants are apportioned asymmetrically prior to cytokinesis, two qualitatively and quantitatively distinct progeny will be generated by division (1, 2). Asymmetric division is a key process for driving cellular diversity in developmental biology (3–7), stem cell biology (4, 8–10) and also recently in the immune system (11–13). Furthermore, molecular asymmetry can involve proteins (4, 6, 11–14), mRNA (5) and DNA (3, 7, 8, 10, 14). The rarity of mitotic events within asynchronously dividing populations necessitates the use of mitotic and cytokinetic inhibitors to obtain statistically relevant numbers of cells for analysis (10, 12, 14). Furthermore, the actual scoring of asymmetric cell division often relies on subjective, non-quantitative imaged-based methods, limited by an inability to directly report cell division history without synchronization or cell sorting techniques. To this end, we wanted to devise a method for studying asymmetric division with the throughput to identify and record high numbers of rare mitotic events without the need for chemical inhibitors. As asymmetric apportionment may only occur within specific division rounds (12, 14) our approach had to also report the division history of the cell population in question without the need for flow sorting or synchronization. To accurately identify mitotic events and objectively measure molecular polarization within these cells we required imagery. While flow cytometry is well suited to analyzing large numbers of cells in an objective, statistically robust manner it does not provided spatial resolution. The recent development of imaging flow technology presents the possibility to study asymmetric cell division and overcome the inherent limitations of imaging and flow cytometric techniques in isolation. In the context of cellular immunity, work by Chang et al argues that effector and memory T cells are generated from a single naïve progenitor during the first round of antigen-triggered division (12). They studied the distribution of several fate determinants during early and late mitosis, including the atypical protein kinase C family member PKCζ (15) that has been shown to be a key player in molecular polarity in T cells (16–18). They found that PKCζ was localized to the synapse-distal daughter pole in ∼70% of telophasic cells undergoing first division, with these daughters correlating with a memory-like phenotype (12). Based on these findings, we have developed an approach combining classical dye dilution techniques with Propidium Iodide (PI) and phospho-histone (pH3) staining that can report the cell division and cell cycle position of all cells within asynchronously dividing populations. We have utilized the Jurkat cell line as a proof-of-concept as it proliferates without stimulus and contains the necessary molecules that are involved in the polarization of primary T cells, including PKCζ and the Par proteins (18). Moreover, it has been previously shown that other polarizing molecules such as flotillins are differentially localized in mitotic Jurkat cells (19). Migrating from a zero resolution flow cytometer to an imaging flow platform preserved the ability to track division while also providing imagery to identify all four stages of mitosis and measure the polarity of PKCζ protein and endosomes within these populations. Importantly, the rapid sample throughput allowed us to collect and analyze statistically relevant numbers of rare events without the need for chemical interventions against the mitotic process that could affect the cells functionality and division capabilities.


Dye Labeling and Cell Culture

Jurkat cells (E6.1, Human Actute T Cell Leukaemia, cultured from an FHCRC derived clone, Cell Services, CRUK,) were maintained in RPMI culture media (Invitrogen, Paisley, UK) containing 10% FBS, Penicillin/Streptomycin, Glutamine and 2-Mercaptoethanol at 37°C/5% CO2. Cells were harvested, counted using a Vi-Cell (Beckman Coulter), washed once in serum-free media and re-suspended in Cell Trace Violet labelling solution (CTV, 2 μM in pre-warmed PBS, Invitrogen, A10198) at a density of 4 × 106/ml at 37°C for 30 min. FBS was added (5% v/v) to absorb cell free CTV. After 10 minutes, cells were washed into serum free media for cell sorting (Supp. Info., Fig. S1). To ensure sufficient CTV labelling uniformity from the input population and subsequent division peak resolution during culture, cells were sorted based on a 40-channel width in the Violet 450/40 detector using a FACS Aria I (Becton Dickinson, BD, Franklin Lakes, USA). Live cells were TOPRO3 negative (Invitrogen). Unsorted or sorted cells were cultured (0.5 × 106 cells/ml) in T75 flasks with complete RPMI. Where indicated, the tubulin inhibitor nocodazole (Sigma St Louis, USA) was added (0.3 μM) for the last 16 hrs of culture. Cells were harvested at defined times then checked for viability (TOPRO-3) and CTV division peak resolution on an LSRII (BD).

Cell Staining

Cells (1 × 106) were seeded in a 96 well round bottom plate, washed once with PBS containing 2% FBS (wash buffer), fixed with 2% formaldehyde (Polysciences, Inc, USA (11814) for 45 minutes at room temperature (RT) then permeabilized with 0.1% Triton-x 100 for 5 min. Samples were stained with either rabbit anti-pH3 Ser 28 (Cell signalling Inc, USA (9713S) and goat anti-PKCζ (Santa Cruz, USA (SC-216-G) or mouse anti-pH3 Ser 10 (Cell signalling Inc, (9706S) and rabbit anti-early endosomal antigen-1 (EEA-1, Santa Cruz, (SC33585) for 1hr at RT followed 2 washes. Cells were incubated sequentially with the appropriate secondary antibodies labeled with either AF488 (for pH3 ser 28/10, anti-rabbit or anti-mouse respectively, both Invitrogen) or AF647 (for PKCζ/EEA-1, anti-goat or anti-rabbit respectively, both Invitrogen) for 45 min at RT in the dark. Cells were washed prior to acquisition on either an LSRII (BD) or an ImageStream-x system (ISx, Amnis corp., Seattle USA). Single stained controls were prepared for spectral compensation.

LSRII Acquisition and Analysis

Samples were incubated with RNAse A (100 μg/ml) and PI (50 μg/ml) for 30 min at RT. CTV (Ex 405nm Em 440/40, 345V), AF488 (Ex 488nm Em 530/30, 423V) PI (Ex 488nm Em 610/20, 469V) and AF647 (Em 633 Ex 660/20, 500V) fluorescence was collected as described. Samples were analyzed using FlowJo software (Tree Star) as outlined in Supporting Information, Figure S2. For cell division analysis, the in-built proliferation platform was used to calculate the percent divided and the proliferative index. The Watson pragmatic fitting algorithm was used for cell cycle analysis. The reported frequencies of G2/M cells from the model fit were adjusted by the pH3+ gated values to give G2 and M percentages.

ImageStream Acquisition and Analysis

Samples (4 × 107/ml in 60μl of wash buffer with 1μg/ml PI and RNAse A) were acquired on a 5-laser 6-Channel ISx Imaging Flow Cytometer with 40× magnification controlled by INSPIRE software and fully ASSIST calibrated (Amnis). Single stained controls were collected with bright-field (BF) illumination off, and with all necessary excitation lasers switched on. Samples were acquired with a BF area lower limit of 35μm2 to eliminate debris. A minimum of 50,000 total events was collected per sample. A compensation matrix was created using single stained raw image files (.rif) and the IDEAS 4.0 compensation wizard (20). The matrix was used to compensate sample .rif files (Supp. Info., Table S1). For brevity, full details of all masking, features and analysis strategies are included in the supplemental notes. Single cell events were identified using the Area and Aspect ratio of Channel 3 (Bright-field) default mask (M03). CTV, PI and pH3 AF488 fluorescence was measured using the Intensity feature and the default M01, M04 and M02 masks respectively. The CTV intensity profile was used to calculate the percent divided and proliferation index as previously described (21). Mitotic cells were identified by PI and pH3 AF488 intensity. Mask M04 was adapted to subdivide pH3+ events using morphology. The Aspect Ratio and Spot Count were calculated for the new mask and plotted as a bi-variate graph to allow population gating. For the identification of telophasic cells and subsequent measurement of signals across the daughter poles, two multi-partite, pole-specific masks were designed. Telophasic events were characterized using a total of nine morphological parameters calculated from the polar masks and their variants (please refer to the supplemental notes for full details).

Method for Determining the Cell Cycle Phase Transit Times

In order to determine the time spent in each of the cell cycle and mitotic phases we used the absolute cell numbers obtained after 96 hours of asynchronous culture. The majority of the cells had undergone three rounds of division, which is consistent with the Jurkat IMT of approximately 24 hours (22). We noted that cells from the fourth division were bunched in the early cycle phases while those from round two were grouped in the later phases. We therefore selected cells that had undergone three rounds of division and were evenly spread throughout their cell cycle to calculate the ratio of cells within each phase and the transition times based on an IMT of 24 hours.


Assessing Cell Division and Cell Cycle by Dye Dilution and PI/pH3

Studying molecular asymmetry in mitotic cells from specific division rounds traditionally requires cell sorting based on fluorescent dye dilution prior to microscopy (12). Reporting cell division and cell cycle position by traditional flow cytometry has been achieved previously using the Hoechst-Brdu quenching method (23). This approach is limited to tracking a maximum of three divisions and is not compatible with staining for other antigens rendering it unsuitable for our needs. To this end we developed a staining panel consisting of CTV, anti-pH3 AF488 and PI. Measuring proliferation by dye dilution requires bright, uniform labelling of the input population or dilution peaks are poorly resolved (24). Unlike primary lymphocytes, Jurkat cells are difficult to label due to non-uniform cell size and dye uptake (25). Therefore we sorted our CTV labeled population prior to culture, selecting a 40 channel width in the Violet 450-50 detector (Supp. Info., Fig. S1) and as such we were able to achieve excellent division peak resolution after 96 hrs of asynchronous culture (Fig. 1A). Nocodazole was used to induce a controlled block in the cell cycle to validate our detection method. Addition during the last 16 hrs of culture had no measurable impact on the percent divided and only reduced the proliferation index marginally (2.39 v 2.16). PI staining allowed us to obtain cell cycle profiles for all four division rounds. Nocodazole treatment led to an increased percentage of cells with G2 DNA content in all rounds of division as measured by PI intensity (Fig. 1B). The inclusion of pH3 AF488 staining revealed that the increase in G2 events in the presence of nocodazole was specifically caused by a 10-fold increase in mitotic cells (Fig. 1C). Interestingly, we noted that mitotic cells possessed a higher PI staining intensity than interphase cells, which was particularly evident in the nocodazole treated samples. It has been suggested that this may be due to differences in DNA condensation during mitosis allowing PI better access to bind (26). Finally, we were able to collate data from multiple experiments and plot the gated frequencies for the four major cell cycle stages within each round of division with or without nocodazole conclusively showing that our approach is able to report the exact position of any single cell within an asynchronously dividing population.

Figure 1.

Measuring the position of cells in the proliferative cycle by dye-dilution, pH3 and PI staining. (A) CTV labeled Jurkat T cells were presorted for bright, uniform dye uptake, and cultured for 96 hrs with or without 0.3 μM nocodazole for the final 16 hrs as indicated. Cells were fixed, permeabilized, and stained for pH3 (AF488) and DNA content (PI) before LSRII acquisition. The undivided peak was determined using serum starved cells. The percent divided and proliferation index is shown for each CTV profile (left panel). Cells (±nocodazole) within each division were identified by CTV dilution and gated as shown (right panel). (B) PI staining was used to determine DNA content and identify G1/0, S-Phase, and G2 populations within each division round (±nocodazole). (C) pH3 staining allowed us to sub-divide the G2/M population into G2 and M events. (D) DNA histograms were subjected to the Watson pragmatic algorithm with G2/M frequencies adjusted for the gated frequencies of pH3+ events to give G2 and M phase values. The mean and SEM of triplicate experiments with (+) and without (−) nocodazole treatment are shown. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary. com.]

Comparing Cell Division Measurements by Conventional and Imaging Flow Cytometry

Next, we migrated our experimental system to the ImageStream platform to obtain the spatial information required to study signal polarisation and asymmetry during mitosis. Although it has been shown previously that signal resolution between the two platforms is comparable using 6-peak fluorescent micro spheres (20), the behaviour of fluorescently stained biological samples may differ. To test the efficacy of population identification and resolution across both platforms we ran samples sequentially. Figure 2A shows that the CTV (M01) dilution peaks were equally well resolved by the ISx and LSRII. The resulting proliferation index values were identical (Fig. 2B). By plotting the PI (M04) intensity and pH3 AF488 (M02) intensity we were able to also identify and gate mitotic events in an analogous manner to LSRII-derived data (Fig. 2C). It should be noted that for ISx analysis PI was not added at a concentration able to achieve stoichiometric DNA measurements due to the danger of saturating the CCD camera. The staining level achieved, while not able to generate satisfactory cell cycle profiles, was able to confirm that pH3+ events had higher DNA content than pH3- events facilitating the comparison of mitotic cell percentages across both platforms. The observed gated frequencies of pH3+ events with and without nocodazole were consistent across platforms (Fig. 2D). Collectively, these data highlight that there was no loss of information or population resolution by migrating our methodology for reporting cell division and cell cycle position to the CCD-based ISx.

Figure 2.

PMT and CCD based flow cytometric systems resolve fluorescently labeled populations equally well. (A) Sample sets were acquired on an LSRII and an ISx system. The CCD-based ISx was able to resolve CTV dilution peaks with the same efficiency as the PMT based LSRII. (B) Kinetic tracking of the proliferation index was also identical across both platforms. (C) pH3 AF488 positivity within the G2 population (±nocodazole) was also equally well resolved (D) the gated frequencies were well correlated across division rounds. The mean and SEM of triplicate experiments with (+) and without (−) nocodazole treatment are shown.

Developing an Imaging Flow Cytometric Approach to Determine Mitotic Phase

The pH3+ population was morphological heterogeneous due to the presence of cells in different stages of mitosis (Fig. 3A). It has been suggested that molecular polarity may be established early in mitosis and stably maintained until cytokinesis (2, 12). To this end we subdivided the pH3+ events into the different mitotic stages using morphological-based parameters. Mask M04 was adapted as outlined in Supporting Information, Figure S4 and supplemental note 2. By calculating the Aspect Ratio and the Spot Count of the modified mask, prophase cells had high Aspect Ratio values and a single bright nuclear spot, metaphase low Aspect Ratio values and a single spot and anaphase cells a range of Aspect Ratio values and two bright nuclear spots (Fig. 3A). As nocodazole blocks at prophase, we were able to use the Aspect Ratio distribution of single spot events from this sample to gate on prophase and metaphase events separately (Fig. 3B). To confirm that metaphase events were not simply anaphase cells imaged in the incorrect 2D orientation, we checked pH3 intensity as anaphase cells have reduced pH3 levels compared to metaphase cells (27). Collating data from multiple experiments allowed us to plot and report the frequencies of prophase, metaphase and anaphase cells within each division round with or without nocodazole (Fig. 3C). As expected, due to inhibition of microtubule formation in the presence of nocodazole preventing metaphase transition the increase in pH3+ events was due to a block at prophase (28). Collectively, these data show that by using morphologically derived fluorescence parameters to analyze pH3+ events collected on an ISx system it is possible to determine the frequencies of prophase, metaphase and anaphase cells. This facilitates the opportunity to conduct a spatial analysis of a given signal within different populations with different division histories.

Figure 3.

Subdividing pH3+ mitotic stages by fluorescence imagery. (A) Adapting the default PI channel mask enabled us to measure the nuclear Aspect Ratio and Spot Count for each event within the pH3+ population. We identified prophase cells (high nuclear aspect ratio with a single spot) metaphase cells (low nuclear aspect ratio with single spot) and anaphase cells (low nuclear aspect ratio with two spots). (B) As an evaluative test for measuring pH3+ mitotic populations by morphology and to set a gate that divided prophase and metaphase events, nocodazole was added in order to introduce a known block at pro-phase. (C) A graph showing the frequencies of morphologically classified prophase, metaphase and anaphase cells (±nocodazole). Values are expressed as the mean ± SEM of three experiments. [Color figure can be viewed in the online issue, which is available at]

Automatic Identification of Telophasic Cells by Morphologically Derived Parameters

Telophasic cells represent the best stage for studying molecular asymmetry as they have clearly defined daughter poles and a cytokinetic plane (10–14). As telophasic cells were pH3- (Fig. 4A lower panel) we used morphologically derived parameters instead. To this end, we calculated and plotted the BF Area and Aspect Ratio and gated on the bi-lobular events (Fig. 4A, upper left panel). This gate contained telophasic cells as well as G1 conjugates and debris (Fig. 4A, image panel). We selected a single 100,000 event file and within the bi-lobular population (2628 events) we manually identified all telophasic cells, mirroring traditional microscopic approaches (10, 12, 14). This population was used to assess the efficacy of each stage in our automated analysis strategy. Designing two multi- partite masks (Supp. Info., Fig. S5), each able to identify the individual nuclear pole of a bi-lobular cell, allowed us to exploit the fact that the nuclear poles of telophasic cells were morphologically distinct to conjugated G1 cells (see Fig. 4A, image panel). These masks also allowed us to measure signal distribution across the daughter poles using a simple custom equation (see supplemental note 3). By performing a detailed appraisal of several pixel-based parameters within the IDEAS software we found that telophasic cells had high Bright Detail Intensity R3 and 2-Fold Symmetry scores in the PI channel (Fig. 4B, top left panel). By calculating and plotting the Aspect Ratio of each tightly masked individual nuclear pole we were then able to eliminate G1 conjugates with one or both spherical poles (Fig. 4B, top right panel). Next, we noted that true telophasic cells had no overlap between the two polar masks we had created and gated out events with significant overlapping area (Fig. 4B, bottom left panel). Finally, we measured the percentage of the CTV and PI signals within both polar masks (Fig. 4B, bottom right panel). We found that bona fide telophasic cells distributed PI within a 45-55% limit and CTV fluorescence within a 40-60% limit, events outside were either a non-telophasic or had been imaged with one daughter pole orientated grossly outside the focal plane and as such could potentially affect asymmetry classification (see Supp. Info., Fig. S6). In normal imaging mode, the ISx has can best resolve objects within a ±2 μm range from the focal point. The effect of subtle defocusing does not alter the total intensity measured but simply spreads the same signal over more pixels (29). Our masking strategy was robust against subtle changes in polar focus with respect to measuring fluorescent intensity for cells within the 40-60% gate. To validate the process we analyzed eighteen data files containing a range of total events. There was an excellent linear correlation between the total event number in the data file and the number of telophasic cells identified (Fig. 4C). We selected four files at random and manually checked to see if we had indeed identified all possible telophasic cells. Our automated image-based analysis pathway was able to identify telophasic events with a 99% (±5%) efficiency rate compared to manual approaches as some were gated out due to gross differences in polar focus. The percentage of telophasic cells was on average 0.05% of all events and when we calculated the phase transition times for all the cell cycle stages from the third division, down to the level of mitosis, we revealed that asynchronously dividing Jurkat cells spend as little as 43 seconds in telophase (Table 1), underpinning the challenge of capturing these events for analysis without chemical intervention. Collectively these data show that we were able to reproducibly and objectively collect and identify statistically relevant numbers of rare telophasic cells using morphologically derived parameters.

Figure 4.

Identifying Telophasic cells using morphologically calculated parameters. (A) Telophasic cells were present at low frequency (3%) within the conjugate population and manually tagged to create a test population (upper left panel, red diamonds). These cells were pH3− (lower left panel). (B) The conjugate population was further refined by plotting the BDI R3 of PI fluorescence (y axis) within the polar masks and the twofold symmetry (x axis) score (upper left panel), followed by the aspect ratio of each pole (upper right panel). Next, the area of overlap between the two polar masks was plotted (lower left panel) followed by the percentage polar distribution of both the PI and CTV signals (lower right panel). (C) Multiple data files (n = 18) containing a range of events were analyzed by plotting the number of telophasic events identified against the total event number in the data file. The r2 and P value for significance from 0 are shown. [Color figure can be viewed in the online issue, which is available at]

Table 1. Cell cycle and mitotic stage transit times
Cell cycle phasePercentage of all cells in each phaseTransit time (min) for IMT = 24 hrs
  1. The absolute cell numbers from data in Figures 1D and 3C allowed us to calculate the phase transition time in minutes for each stage of the cell cycle, down to the level of mitotic phase. The errors calculated for the percentage of cells in each phase are derived from the standard deviations of the cell numbers in each phase (and the total cell count) of three experiments.

G0/148.6 ± 1.1700.2 ± 15.6
S41.3 ± 0.8595.2 ± 11.4
G28.4 ± 0.3121.8 ± 4.2
M1.6 ± 0.123.4 ± 1.2
Prophase1.064 ± 0.23415.32 ± 3.37
Metaphase0.468 ± 0.1466.75 ± 2.10
Anaphase0.018 ± 0.0160.26 ± 0.23
Telophase0.049 ± 0.0070.70 ± 0.10

Detecting Asymmetric Distribution of Fate Determinants Using Imaging Flow Cytometry

Having designed and validated an approach to identify cells in all four mitotic stages we looked at the distribution of PKCζ protein by immunofluorescence in mitotic Jurkat cells. The 40-60% gated limits for PKCζ fluorescence were set using the distribution of CTV to account for potential inherent asymmetry of proteins previously described in post-mitotic cells (30). As Jurkat cells are immortalized cells and were not cultured under polarising conditions we did not expect to observe any PKCζ asymmetry outside our 40-60% limit. However it should be noted that under conditions of homeostatic proliferation Chan et al did note PKCζ asymmetry (12). Interestingly we noted that 2.8% of telophasic cells distributed PKCζ fluorescence asymmetrically outside the 40-60% gate while still maintaining CTV fluorescence within these boundaries. CTV dye dilution showed that PKCζ asymmetry occurred in cells entering first and second divisions (data not shown), although we would not expect Jurkat cells to show any divisional bias unlike primary T cells (12). It has been reported that PKCζ polarity is stably maintained across mitosis (12). As there are no clear daughter poles to define the plane of symmetry in prophase and metaphase we used a different approach to measure polarity at these stages. We measured the delta between the intensity-weighted centroid of the PKCζ image and the non-intensity weighted centroid of the BF image and compared this distribution to the delta between the centroids of the CTV and BF images as a gating control for little or no polarity (see supplemental note 2). Using this approach we could detect PKCζ signal polarity within prophase (Fig. 5B) and metaphase cells (Fig. 5C). We also measured the polarity of EEA-1 labeled endosomes within different mitotic stages as a known control for symmetrical protein inheritance. This was particularly important as it would also control for any potential noise in our methodology. Interestingly we found that EEA-1 asymmetry in telophasic events was significantly lower compared to PKCζ with polarity at early mitotic phases not translating well to telophase (Fig. 5D). Collectively these data show that by using intensity based spatial parameters, we can measure the polarity of a given signal within mitotic cells in a controlled, non-subjective manner.

Figure 5.

PKCζ distribution in telophasic cells is significantly more asymmetrical than EEA-1 labeled endosomes. (A) The percentage of total PKCζ signal distributed within each daughter pole of telophasic cells (upper left panel) was plotted (x axis) versus the percentage of the CTV signal within the same (y axis). Symmetrical events are defined as lying within the 40 to 60% limit of the CTV and the PKCζ distributions. Asymmetrical events also fall within the 40 to 60% of CTV distribution but within either the 0 to 40% or 60 to 100% gated areas for the PKCζ distribution. Multispectral single and merged images of telophasic cells with nonpolarized (upper right panel) and polarized (lower right panel) PKCζ signals are shown. Polarity measured using Delta-centroid x, y values between the BF image and intensity weighted AF647 Image (open histogram) for prophasic (B) and metaphasic (C) pH3+ gated events. The polarity gate was set based on the delta centroid of the BF and CTV image pairs (filled histogram). (D) A graph showing the percentage of events with polarized PKCζ or EEA-1staining across the mitotic stages. Values are expressed as the mean ± SEM of seven independent experiments. The total number of telophasic events analyzed for PKCζ and EEA-1 distribution were 434 and 446, respectively. Significance was determined using student's t-test (*P = 0.02, and ****P < 0.0001). [Color figure can be viewed in the online issue, which is available at]


We have described a novel approach to studying molecular asymmetry during mitosis using imaging flow cytometry. Transferring experiments from a traditional PMT-based system to the CCD-based ISx preserved population resolution allowing us to report division history and cell cycle position for each event acquired to the mitotic stage level. This overcame the need to sort (12) or synchronise the cells (14) to subsequently analyze mitotic events from specific division rounds by microscopy. The throughput of this approach was particularly important as Jurkat cells only spent 43 seconds in telophase, representing no more than 0.05% of all events within an asynchronously dividing population. This eliminated the need for chemical interventions to artificially enrich target cells as it hard to determine what effect inhibitors may have at the molecular level, particularly as cytochalasins have been shown to affect both calcium and phosphoinosotide metabolism in leukocytes (31). Analyses of rare cell populations by zero-resolution flow cytometry are governed by guidelines driven by the absence of imagery. These include how many total events need to be collected based on the target frequency as well as the various staining controls required for robust CVs for the final measurement (32, 33). The observed frequency of telophasic events (0.05%) was within the boundaries of rare cell analysis but the spatial information afforded by imaging flow cytometry allowed us to easily conclude if an event was real or not. Furthermore, we used a total of nine separate morphologically-derived parameters to identify telophasic cells within large data sets as well as two internal staining controls to set the boundaries for any inherent post-mitotic asymmetry (30). Manually scoring asymmetry using multi-colour microscopy often requires varying collection filters and exposure times, which can create issues with spatial registration and the relative signal quantification across image channels. The unified fluorochrome excitation and emission capture by the ISx combined with sub pixel spatial registration of the spectrally decomposed images means that cross-channel image analysis can be performed with confidence. Collectively, these considerations eliminated the subjective nature of a manual mitotic identification process, allowing us to determine signal polarity across the mitotic phases in a quantitative manner. In our study we noted ∼3% of 434 verified telophasic cells exhibited PKCζ asymmetry and under homeostatic proliferation conditions Chang et al reported 14% of 29 manually identified telophasic had polarized PKCζ fluorescence. In our study asymmetry, while seemingly striking by eye, was subtle when the relative polar distribution was quantified by the ISx CCD array. This highlights further limitations of manual microscopic approaches as quantitative decisions based on human interpretation may incur error. One could speculate that the manual scoring approach used by Chang et al could have over reported asymmetry within the homeostatic and antigen activated cell populations. From an immunological perspective it makes little sense if during infection 70% of naïve T cells, albeit within the first mitosis, asymmetrically divide to generate one effector and one memory cell. One could argue that the incidence of asymmetry should be far lower to preferentially generate effector cells during the acute infection phase to clear pathogen, with only a few memory cells required for protective immunity against delayed challenge. We are currently using our approach to re-evaluate these findings using mouse TCR-transgenic systems. A recent report using elegant fluorescence bar-coding techniques to track single cell fates in the immune system did not preclude asymmetric division in the immune response (34). It is beyond the scope of this paper to prove if asymmetry really occurs during an immune response, we simply wanted to provide a technical and analytical framework that overcomes the limitations of traditional approaches. Although our proof-of-principle system utilized an immortalized cells line, our data shows that there was a low incidence of asymmetric PKCζ distribution even though this would have no influence on cellular fate. It has been shown previously that there is a generalized asymmetry of proteins in post-mitotic cells (30). To this end we used the CTV-labeled cellular signal as the gating control to set the limits of accepted variation in protein distribution. The fact that in certain cells PKCζ distribution is specifically outside these limits is key. Asymmetric division has previously been described in immortalized cell lines with respect to the inheritance of BrdU labeled template DNA (14). As a further control we measured EEA-1+ endosomal distribution across the mitotic plane as this has been shown to follow a random, symmetrical distribution (35). As expected we found that EEA-1 distribution was significantly less asymmetrical compared to PKCζ protein in telophasic events. Of note, it has been suggested that when endosomes are loaded with nano-particles, they may be inherited in an asymmetric fashion (36). Using our approach, we were able to identify sufficient numbers of prophase and metaphase cells to address the question of early mitotic stage polarity and subsequent translation of asymmetry to telophase. We found that the percentage of prophase and metaphase cells with polarized PKCζ did not differ significantly, but did decrease significantly at telophase. Furthermore, endosomal polarity was also higher at prophase but dropped at metaphase and telophase. Polarisation of endosomes at prophase has been described previously and may be due to association with the MTOC (35). These data suggest that early mitotic phase polarity does not translate to true asymmetry at more progressive stages where a plane of symmetry can be defined. Interestingly, Chang et al show that PKCζ polarity is established early in prophase and translates stably to telophase. Our analysis approach and masking strategies can easily be applied to ex-vivo primary mouse and human immune cells to determine the in vivo asymmetric responses to infection or homeostatic conditions; these studies are currently in progress. One of the main challenges faced when looking at primary cells is that the transition time for telophase is reduced making these events even harder to capture. However we have found that the throughput afforded by imaging flow is able to circumvent this issue. There is some debate as to whether asymmetric division occurs in specific settings (37) and it may be interesting to take our approach and reassess asymmetry in these situations. It may also be valid to look at the apportionment of key cellular structures such as mitochondria. Finally, our system is not limited to study asymmetric division; we can also look at cell cycle and mitosis transition times during drug treatment and determine the effects on cells with different proliferative histories.


The authors also acknowledge the help and advice of our colleagues in the LRI and CRI CRUK Flow Cytometry Laboratories. This study was conducted from June 2009 till January 2011.