A critical appraisal of factors affecting the accuracy of results obtained when using flow cytometry in stem cell investigations: Where do you put your gates?

Authors

  • Owen R. Hughes,

    1. Institute of Human Genetics and North East England Stem Cell Institute, University of Newcastle, International Centre for Life, Newcastle upon Tyne, United Kingdom
    Current affiliation:
    1. MRC Centre for Regenerative Medicine, University of Edinburgh, Edinburgh, United Kingdom
    Search for more papers by this author
  • Rebecca Stewart,

    1. Institute of Human Genetics and North East England Stem Cell Institute, University of Newcastle, International Centre for Life, Newcastle upon Tyne, United Kingdom
    Search for more papers by this author
  • Ian Dimmick,

    Corresponding author
    1. Institute of Human Genetics and North East England Stem Cell Institute, University of Newcastle, International Centre for Life, Newcastle upon Tyne, United Kingdom
    • Institute of Human Genetics, University of Newcastle, Bioscience Centre, International Centre for Life, Central Parkway, Newcastle upon Tyne, NE1 3BZ United Kingdom
    Search for more papers by this author
    • Ian Dimmick and Elizabeth A. Jones contributed equally to this work.

  • Elizabeth A. Jones

    1. Institute of Human Genetics and North East England Stem Cell Institute, University of Newcastle, International Centre for Life, Newcastle upon Tyne, United Kingdom
    Current affiliation:
    1. Genetic Medicine, Manchester Academic Health Science Centre, Central Manchester University Hospitals NHS Foundation Trust, Oxford Road, Manchester M13 9WL, United Kingdom
    Search for more papers by this author
    • Ian Dimmick and Elizabeth A. Jones contributed equally to this work.


Abstract

Flow cytometry is used extensively in stem cell investigations but there is wide variation in the methods used for data analysis between laboratories. Data analysis can be challenging in stem cell biology as there is often no clear distinction between positive and negative populations. We have undertaken a critical appraisal of factors that affect the accuracy of results in stem cell applications. We used mouse embryonic stem cells and determined the expression of three common antigens in stem cell investigations, namely CD15, CD184, and c-kit. We have compared different cell preparation methods and gating strategies and also evaluated the use of isotype controls and unstained cells as controls for the identification of positive populations. The use of a “doublet discriminator” using a side scatter area signal versus side scatter height signal dot plot to identify single cells for analysis increases the accuracy of results regardless of the method used to dissociate cells. Isotype controls can be helpful in mimicking cellular nonspecific binding of the experimental antibody reaction. Isotype controls behave differently on stem cells at different stages of differentiation. Analysis of a viable single cell population with careful selection of control cell populations increases the accuracy of results. © 2009 International Society for Advancement of Cytometry

There has been increasing interest in stem cells due to their potential uses in regenerative medicine (1). Stem cells share the unique characteristic of self renewal and the ability to differentiate into several cell types and so provide a potentially unlimited source of cells for transplantation. In addition, by using stem cells to study the process of differentiation in vitro, it is possible to gain insights into the mechanisms of development (2).

There are several different types of stem cell such as embryonic stem (ES) cells, cord blood stem cells, adult stem cells such as haematopoietic stem cells, and induced pluripotency stem cells (3). These cells share many common features but other properties such as their cell surface marker expression differ widely. Human hematopoietic stem cells (HSCs) can be obtained from a variety of hematopoietic tissues, including bone marrow, blood, cord blood, and fetal liver (4–7). ES cells are derived from the inner cell mass of the blastocyst (8–10). Mouse ES cell lines can be derived by culturing the inner cell mass cells of mouse blastocysts on embryonic fibroblasts in the presence of leukemia inhibitory factor to prevent spontaneous differentiation.

Flow cytometry has become an indispensible tool in stem cell biology. It is used for both the phenotyping of cells and sorting of cells into different populations based on their physical properties and protein marker expression profiles. Accurate identification of these populations is crucial. Many of the techniques in flow cytometry were developed using mature leucocytes, predominantly lymphocytes and these cells are a nonadherent cell type. This technology has been adapted for use in stem cell biology. However, stem cells have some unique characteristics and so some refinements are required to optimize the technique. Specific gating protocols exist for diagnostic and clinical purposes especially in relation to applications in haematology (11, 12) but few standard protocols have been developed for other stem cell applications.

Flow cytometry can be defined as a semi automated procedure for the interrogation of single cells in a continuous fluid stream enabling the derivation of simultaneous measurements of multiple extra- and intracellular characteristics by the use of appropriate probes. The objective of flow cytometry is very simple; to measure, by quantifying photon release, constituents of the membrane, cytoplasm and nucleus of a particular cell or group of cells. Fluorescently labeled antibodies raised against cell surface and intracellular markers can be used to identify cell populations of interest. The use of gating strategies in flow cytometry is a highly important process for selecting populations of interest, but also to ensure that data is correctly and accurately analyzed. The best way to distinguish between fluorescent positive and fluorescent negative populations has been the subject of debate (13, 14). Some specialists advocate the use of isotype controls, some isoclonic controls, and others the use of unstained cells alone.

Isotype controls are an antibody of the same immunoglobulin subclass and from the same species as the primary antibody but raised against an antigen that is not usually found in the biological sample of interest. It is used to confirm that the primary antibody binding is specific and not a result of nonspecific Fc receptor binding or other protein interactions. Isotype controls should be matched to the immunoglobulin subtype and used at a similar protein concentration as the test antibody as well as being conjugated to the same fluorochrome. Isotype controls do not have the identical tertiary structure of the test antibody as the antigen specificities are different. The nonspecific binding, and therefore, the degree of fluorescence added to the negative cells by nonspecific/Fc binding may not be identical to that of the test antibody, and this is one reason why the use of isotype controls can be misleading. The best nonspecific binding control is the negative population within the positive sample, but this is only useful if there is a clear demarcation between negative and positive populations. In this case, it is possible to see exactly the nonspecific/Fc binding attributable to the test antibody on the negative population (15). The correct control is of crucial importance when analyzing cells that do not clearly resolve into positive and negative subsets. In our experience, this is a particular problem in ES cell investigations.

In this article, we have concentrated on the use of murine ES cells and the challenges they pose. We have also used blood and bone marrow cells to illustrate a comparison between cell types. However, the principles that we illustrate here will have wide application in investigations using many different stem cell types as identification of specific cell types is a common aim and has been the subject of a recent excellent focus issue in this journal (16).

We have utilized antibodies of different immunoglobulin subtypes and conjugated to different fluorophores. Each of these antibodies is well characterized and widely used within the field of stem cell research. CD15 (SSEA1) is expressed on granulocytes, monocytes, and undifferentiated murine ES cells (PMID: 11278338 and PMID: 281705). We have used an IgG3 anti-CD15 monoclonal antibody conjugated with fluorescein isothiocyanate (FITC). CXCR4 (CD184) is a 7-transmembrane span, G-protein coupled receptor and is expressed on a broad range of cells types including haematopoietic cells and ES cells (PMID: 17169327). We have used a rat IgG2b, κ anti-CXCR4 monoclonal antibody conjugated to R-phycoerythryn (PE). C-kit (CD117) is a transmembrane tyrosine kinase receptor expressed on haematopoietic precursors (PMID: 1711568). We have used a rat IgG2b, κ anti-CD117 monoclonal antibody conjugated to allophycocyanin (APC).

This article details some of the factors that we have shown to be important when conducting investigations using ES cells, and we conclude by outlining a cell population identification strategy for use in these investigations.

MATERIALS AND METHODS

Preparation of Lymphocytes and Bone Marrow Cell Samples

A human blood sample was taken from an adult volunteer, and EDTA used as the anticoagulant. In accordance with the ethical 3R's principle, adult C57/B6 mice that were being culled according to Home Office guidelines were utilized so as to prevent unnecessary additional sacrifice. The mice were sacrificed, and blood was aspirated from the thoracic cavity. The femurs were dissected and flushed with phosphate buffered saline and the bone marrow cells collected. Erythrocyte lysis was performed using a working solution of Pharmlyse (BD-Pharmingen) at a ratio of 50 μl of blood to 1 ml of Pharmlyse. This was then incubated at room temperature for 30 min. The cells were then washed in PBS and reconstituted to a total volume of 300 μl.

Embryonic Stem Cell Culture

E14 mouse ES cells were cultured under serum free conditions according to the manufacturers instructions (ESGRO Complete™, Millipore). The karyotype was checked at regular intervals and a normal karyotype maintained. Cells were routinely tested for mycoplasma contamination. Low density plating assays were performed to ensure the cells were being maintained in an undifferentiated state. The majority of colonies showed no differentiation after alkaline phosphatase staining. To obtain differentiated cells, the ES cells were plated onto gelatine-coated wells in a 12-well plate at a density of 20,000 cells per well in 2 ml of ESGRO Complete™ clonal grade media. After 24 h, the media was changed to ESGRO Complete™ basal media. Media was changed at least every 48 h. After 6 days in ESGRO Complete™ basal media, the cells were prepared for staining.

Preparation of Mouse ES Cells for Staining

The cells were grown in 0.1% gelatine-coated T25 flasks (or 12 well plates). The media was aspirated, and the cells washed in 10 ml of PBS. ESGRO Complete™ accutase (1 ml), ESGRO Complete™ trypsin, or TVP (0.025% trypsin, 1% chicken serum 0.1 mM N2EDTA) were added, and the cells incubated at 37°C until the cells were detached. ESGRO Complete™ basal media (5 ml) was then added and the cell suspension centrifuged at 1000 rpm for 3 min. The pellet was resuspended in PBS, centrifuged again, and then resuspended in PBS before cell counting.

Cell Counting

To obtain accurate cell counting before antibody staining we used the Beckman Coulter Vi Cell XR (Beckman Coulter, High Wycombe, UK). Viability was also estimated using this instrument by Trypan Blue exclusion.

Cell Staining

The antibodies used were as follows: anti CD15-FITC (Cytognos—CYT-15F4), anti CD117(c-kit)—APC (BD Pharmingen™ 553356), anti CD184(CXCR4)—PE (BD Pharmingen™ 551966), mouse IgG3-FITC isotype control (Cytognos—CYT-IC010F-25), PE rat IgG2b isotype control (BD Pharmingen™ 553989), APC Rat IgG2b isotype control (BD Pharmingen™ 553991). All antibodies were titrated on the appropriate cell type before use. We optimized the antibody concentration to the cell count. All isotypes were Ig subtype matched and used at the same protein concentration as the corresponding antibody. The correct volume of antibody or isotype was then added to 100,000 cells in 300 μl PBS and incubated for 20 min at RT. The cells were then washed using the wash only cycle of the Becton Dickinson lyse wash assistant. DAPI (50 μl) was added to each sample before analysis.

Cell Analysis

Cells were analyzed in the FACSCanto II flow cytometer (Becton Dickinson, San Jose, CA). The instrument quality control was checked on a daily basis throughout this study. This was performed using Cytometer setting and tracking (CS&T) beads and software (Becton Dickinson) to check for instrument fluorescence intensity drift and linearity and to adjust laser delay settings and area scaling factors. Spherotech ultra fluorescent particles (Spherotech, Lake Forest, IL) were also used following the CS&T beads to monitor the consistency of measurement of these beads for a known position on a linear scale (channel 50,000) under constant PMT voltages for each detector as a second check to verify instrument precision from day to day. The instrument configuration was the standard 3 laser 488-nm solid state, 20-mW, detectors (Blue: 530/30; 585/42; >670; 780/60 nm) 405-nm solid state diode, detectors (Violet: 450/50; 502 to 525 nm) 30-mW and 633-nm HeNe, 17-mW detectors (Red: 660/20; 780/60 nm). The PMT settings for each individual cell type were achieved by modifying the CS&T baseline voltages so as to achieve the minimal CV for each unstained sample and optimizing signal to noise.

A primary gate was placed on the scatter population and also the DAPI negative cells enabling a Boolean gate so to analyze only live cells. In addition, we incorporated a doublet discriminating gate based upon height versus area of the side scatter signals. The final gate for analysis was a Boolean gate of scatter, live cells, and singlet gated cells. Baseline photomultiplier settings were established using unstained cells from each particular sample group. These settings were then used throughout the assay for that sample group. Histogram gates were set with a 0.5% cutoff.

Statistical Analyses

All graphical error bars represent the standard error of mean measurement values from biological replicate samples. All P-values were calculated using an unpaired, two-tailed Student's t-test.

RESULTS

The accuracy of results determined by flow cytometry can be broadly split into antibody dependant and independent factors. Antibody independent factors include instrument factors such as the accurate setting of PMT voltages and the monitoring of the instrument for stability, bR and qR values and the calibration of area scaling factors. It is generally accepted that use of a viability dye to exclude dead cells from analysis is good practice and all the data shown in this article are derived from measurements of the live population of cells.

Stem cells are often grown adherent to a tissue culture vessel or in three-dimensional structures called embryoid bodies. In both cases, a single cell suspension must be formed by enzymatic dissociation before analysis. Cells can be brought into suspension by enzymes such as trypsin and accutase. Trypsin is a pancreatic serine protease and accutase combines protease and collagenolytic activities. However, enzymatic dissociation can affect the viability of cells and so a balance must be struck between the maintenance of cell viability and the degree of separation. Thus, doublets and clumps are likely to be present in the resulting cell suspension.

To identify a single cell population for analysis, we have used a side scatter area and side scatter height plot (17). Discrimination of cell doublets and cell size measurement by pulse-with analysis was introduced by Tom Sharpless in the early 70's (18). Doublets and clumps cannot always be reliably identified on a standard forward scatter area versus side scatter area plot, and so, if a gate is drawn on this plot, there is a chance that doublets and small clumps will be included in the analysis gate. The area and height of any signal should be roughly proportionate when a single cell is analyzed. The basis of this lies with the normal distribution of photon release from a single cell under transient laser interrogation (Fig. 1). When cells become clumped, the interrogation of the cell clump yields a pattern of increased side scatter area when viewed in proportion to side scatter height. Therefore, if a side scatter area and side scatter height plot is used, doublets and clumps can be identified and excluded from the analysis gate. This is important because if the whole stem cell population is considered, omitting this doublet discrimination, the resultant data would be an inaccurate representation of single cell interrogation in proportion to the degree of clumping. The individual cells within the clump will not be able to be evaluated accurately for any chosen parameter.

Figure 1.

Doublet discrimination. A theoretical representation of the relative position of cells being interrogated by a 488 nm laser beam, with respect to side scatter area and side scatter height. (A) For single cells the measurements from each event are plotted and are proportional. (B) When two cells are adherent to each other, the area signal becomes disproportionate to the height signal. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com]

To ascertain how this affects data analysis we analyzed an unstained cell population derived from a mES cell line dissociated with accutase and compared it with a human blood cell population (Fig. 2). In human blood, clumping is not evident and so a gate can be accurately placed with respect to side scatter. However, in the mES cell suspension, some doublets and clumps were present. The single cell population can be identified on the side scatter area versus side scatter height plot. When this single cell population is considered for analysis, the gate is placed at lower fluorescence intensity than if all events are considered. This change in gate placement affects the accuracy of results.

Figure 2.

Fluorescence differences between single and clumped cells. A comparison between mouse ES cells (mES) and human blood cells (hblood) is shown. (A) and (F) unstained cell suspensions illustrating the doublets and clumps present in the mES cell suspension. (B) and (G) SSC-A versus FSC-A dot plots showing the position of the primary gate. (C) and (H) fluorescence histograms (1-633/660/20-A) of populations (P1) (D) and (I) SSC-A versus SSC-H dot plots of P1 populations enable the discrimination of clumped cells present within the mES sample (D) and reveal the absence of clumped cells within the hblood sample (I). Inserts a–c are expanded views of relevant regions of these populations. (E) Fluorescence histogram displaying single and clumped mES cell populations separately and the difference in gate position for each population. Boxed numbers within C and E represent the channel numbers of corresponding gates. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com]

To determine if the method of cell separation could be optimized to minimize this issue, we undertook a comparison of three common methods of dissociation: accutase, trypsin and trypsin, versene and chick plasma (TVP). Once dissociated, DAPI was added to triplicate samples of 100,000 cells for each method of dissociation. From each sample 10,000 cells were analyzed in the FASCanto to estimate cell viability and the percentage of single cells. In our hands, there was no statistical difference (unpaired, two tailed t-test), between the methods (Fig. 3). In all three cases, complete dissociation of clumps was not achieved without unacceptable levels of cell death. However, in routine use, we found that accutase reliably generated sufficient cells of interest (single and viable), and therefore, accutase was used to separate cells for antigen determination in the subsequent investigations.

Figure 3.

Comparison of the efficacies of three cell dissociating enzymes. Graphical representation of the percentage of single cells (determined using SSC-A vs SSC-H dot plots), and the percentage of these which are live (defined by an ability to exclude DAPI) upon the dissociation of mES cells using Accutase, Trypsin, or TVP. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com]

The accuracy of results determined by flow cytometry is also affected by antibody dependant factors. Each antibody is raised against an antigen of interest. It is essential to titrate the antibody being used to an optimum amount for a set number of cells in a given volume to ensure accuracy of results by maximizing the differentiation of positive and negative events and comparability between experiments. This has been done for all the antibodies shown here. The correct titration of antibody maximizes the signal by saturating the antigen and minimizes background.

The final tertiary structure of the antibody will be determined by its immunoglobulin subtype and the antigen target. Thus, the nonspecific binding of any antibody will vary. When the fluorescence intensities of the positive and negative populations overlap, it is essential to try and mimic the nonspecific binding of an antibody in order to correctly determine the percentage of positive cells in a sample. This is most commonly achieved by using an unstained cell population or an isotype control. We have compared the results obtained using three different antibodies conjugated to different fluorochromes.

IgG3 Anti-CD15 Monoclonal Antibody Conjugated with FITC

The isotype control has a low level of nonspecific binding and so the results when the unstained cell population or cells stained with a FITC conjugated isotype are not significantly different when undifferentiated mES cells or human blood cells are analyzed (Fig. 4). The CD15 staining of human blood cells illustrates the clear resolution between positive and negative cell populations. This is not the case when mouse ES cells are stained with the same antibody. Here, the two populations have a continuous distribution. However, the use of an isotype control enables accurate gate placement the percentage of CD15 positive cells can be accurately determined.

Figure 4.

CD15 analysis of undifferentiated mES cells and human blood cells. Fluorescence histograms (2-488/530/30-A) of single and viable cells from (A) unstained undifferentiated mES cells (mES), (B) unstained human leukocytes, (C and D) isotype control (IgG3) stained cells, and (E and F) CD15-FITC stained cells. Alternative gates for the identification of CD15 expressing cells within CD15-FITC stained samples were generated using both the unstained and the isotype control stained samples. Boxed numbers within A–D represent the channel numbers of corresponding gates. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com]

IgG2b, κ Anti-CXCR4 Monoclonal Antibody Conjugated to PE

There is no clear separation between CD184 positive and negative populations regardless of the cell type being analyzed (Fig. 5). The PE conjugated isotype has a much higher level of nonspecific binding than was seen for the FITC isotype. Interestingly, the nonspecific binding is higher in undifferentiated cells than in more differentiated cells and this increases the possible overestimation of antigen expression.

Figure 5.

CD184 analysis of undifferentiated mES cells, differentiated mES cells, and mouse blood cells. Fluorescence histograms (2-488/585/42-A) of single and viable cells from (A) unstained undifferentiated mES cells (mES), (B) day 6 differentiated mES cells and (C) mouse blood cells (mblood), (D, E, and F) isotype control stained cells and (G, H, and I) CD184-PE stained cells. Alternative gates for the identification of CD184 expressing cells within CD184-PE stained samples were generated using both the unstained and the isotype control stained samples. Boxed numbers within A–F represent the channel numbers of corresponding gates. (J) The potential difference in the estimation of CD184 positive cells is shown as the mean difference (percentage using unstained gate − percentage using isotype control gate). This has been calculated for cells appearing to express CD184 based upon application of the alternative gates to CD184-PE stained samples. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com]

IgG2b, κ Anti-CD117 Monoclonal Antibody Conjugated to APC

Again, there is no clear separation between c-kit positive and negative populations (Fig. 6). The APC conjugated isotype has a very high level of nonspecific binding regardless of cell type. However, the potential overestimation of c-kit expression remains highest for the undifferentiated ES cells. The lack of separation between positive and negative populations highlights the need for some estimate of nonspecific antibody binding.

Figure 6.

C-kit analysis of undifferentiated mES cells, differentiated mES cells and mouse bone marrow cells. Fluorescence histograms (1-633/660/20-A) of single and viable cells from (A) unstained undifferentiated mES cells (mES), (B) day 6 differentiated mES cells and (C) mouse bone marrow cells (mbone marrow), (D, E and F) isotype control stained cells and (G, H and I) c-kit-APC stained cells. Alternative gates for the identification of c-kit expressing cells within c-kit-APC stained samples were generated using both the unstained and the isotype control stained samples. Boxed numbers within A–F represent the channel numbers of corresponding gates. (J) The potential difference in the estimation of c-kit positive cells is shown as the mean difference (percentage using unstained gate − percentage using isotype control gate). This has been calculated for cells appearing to express c-kit based upon application of the alternative gates to c-kit-APC stained samples. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com]

DISCUSSION

Flow cytometry is an invaluable tool in stem cell biology as evidenced by its use in an increasing number of applications (16). The accurate identification of cell populations of interest is crucial. We have shown that it is advantageous to analyze antigen expression on a population of single, viable cells, and this reduced possible overestimation of antigen expression caused by clumped cells. We suggest that utilization of a viability dye in conjunction with the side scatter area versus side scatter height dot plot to enable a “doublet discriminator” gate would be a helpful addition to all analyses of cell populations that may have some cell doublets or clumps present. In addition, care should be taken when working up appropriate controls.

Isotype controls are useful in mimicking the nonspecific binding of antibody to the sample cells. However, the binding of isotype varies between cell types and their stage of differentiation so control and sample cell types must be carefully matched. If isotype control data is not considered when analyzing results, there is a risk of overestimating the percentage of positive cells. However, as isotype structure does not exactly match the tertiary structure of the antibody of interest (as they are raised to different antigens), they are an imperfect estimate of nonspecific binding. We suggest that an appropriate isotype control be used if a suitable one is available. However, as our data has shown this is not always possible and in these circumstances it may be necessary to rely on unstained cells as an imperfect substitute.

An additional factor that must be considered in stem cell investigations is the use of multiple fluorochromes. Although a full analysis of the choice of fluorochrome combinations and their spectral overlaps is beyond the scope of this article, it is important to ensure correct compensation by the use of good and appropriate compensation controls when placing gates in this situation.

We suggest that all authors publish details of the gating strategies and the control cell populations they have used as this would aid comparison of results from different investigations. Our aide memoire for selecting the cell population of interest is as follows:

  • 1Set up—ensure FACS machine is optimally quality controlled
  • 2Employ compensation to the sample if required
  • 3Live cells—use a viability dye
  • 4Exclude clumps and doublets using the doublet discriminator
  • 5Controls—use isotype controls where possible and appropriate and cell type specific control cells
  • 6Target the cell population of interest

Acknowledgements

The authors are grateful to Paul Cairns for his assistance in the preparation of mouse cell samples and the Prenatal Section of the Cytogenetics Laboratories of the Northern Genetics Service for karyotyping the mES cell lines.

Ancillary