Flow-Based Combinatorial Antibody Profiling Protocol

Materials of Flow-Based Combinatorial Antibody Profiling Protocol

Supplies and Equipment

1. Lyophilized BD FACS™ CAP plates (see Note 1).
2. Cytometer: A cytometer equipped with a 488-nm laser and a 633-nm laser, and the appropriate filter sets to detect FITC/Alexa Fluor 488, PE, and APC/Alexa Fluor 647 are required. In addition, the cytometer must be equipped with a HTS capable of running 96-well plates. This technology has been used successfully on the BD FACS™ Calibur equipped with CellQuest software, BD FACS™ Canto equipped with FACSDiva™ software, and BD LSRII equipped with FACSDiva™ software.

Reagents and Cells

1. Antibodies: All monoclonal antibodies are directly conjugated to fluorescein isothiocyanate (FITC), Alexa Fluor™ 488, R-phycoerythrin (PE), allophycocyanin (APC), or Alexa Fluor™ 647. The antibodies are formulated as three-color cocktails in 79 wells of a 96-well plate at experimentally defined saturating concentrations.
2. Human FcR blocking reagent.
3. Wash buffer and stain buffer.
4. 1% paraformaldehyde.
5. BD CompBeads™ Plus (see Note 2).
6. (Optional) LIVE/DEAD® fixable dead cell stain kit (see Note 3).
7. Any cell type that is amenable to flow cytometry may be analyzed by BD FACS™ CAP.

Analysis Software

Custom scripts written using the R software (1) were utilized for analysis. R is an Open Source environment for the implementation of statistical computing and graphics. It provides a large, well-integrated collection of tools for data management, data storage, data analysis, and graphical displays. There is a very large community of R users that provide custom packages to address specialized problems such as those discussed here. R includes a very flexible and effective programming language for automating many complex procedures, which provides the basis for our custom scripts. This combination of specialized analysis packages and custom scripts allows for a highly specialized, robust, efficient, semiautomated approach to flow cytometry data analysis.

Methods of Flow-Based Combinatorial Antibody Profiling Protocol

Staining Protocol

1. Harvest the cells to be stained and determine cell number and viability. Ensure that the viability is high in each sample to be processed. If the samples have low viability, a live/dead cell discrimination dye may be added to the staining protocol to minimize the false-positive rate.
2. Resuspend cells to a concentration of 10⁷ cells per 90 mL of stain buffer.
3. Add 10 mL of FcR blocking reagent per 10⁷ cells, mix well, and incubate at 4°C for 10 min.
4. Dilute the cell suspension with BD stain buffer to a concentration of 15 × 10⁶ cells/30 mL so that each 100 mL contains 5 × 10⁴ cells.
5. Add 100 mL of this cell suspension to the staining wells of the 96-well plate containing the lyophilized antibody cocktails.
6. Add cells to the empty wells for assay setup and preparation of compensation controls.
7. The compensation markers should be chosen according to the specific cell type that is being analyzed.
8. Incubate the plates at 4°C for 30 min.
9. Add 100 mL of BD wash buffer to each well, centrifuge at 300 × g for 5 min, and decant the supernatant.
10. Add an additional 250 mL of BD stain buffer to each well, mix, centrifuge at 300 × g for 5 min, and decant the supernatant.
11. Resuspend the stained cells in 200 mL of freshly prepared 1% paraformaldehyde.
12. Store the plates at 4°C in the dark until acquisition.
13. Data should be acquired within 24 h after fixation.

Data Acquisition

Data can be acquired on a BD FACSCanto™ or BD LSRII flow cytometer equipped with a HTS. Data acquisition is performed with BD FACSDiva™ software.

1. Create a gate on a dot plot of forward scatter (FSC) versus side scatter (SSC) defining the viable cell population of interest.
2. Collect a minimum of 2,000 events per well.
3. During acquisition, closely monitor histogram plots of all the relevant fluorochromes to ensure that no anomalous events occur.
4. If a viability dye is used, first gate on the live cells to ensure that only viable cells are plotted on the FSC versus SSC plot described above.

Data Analysis

Successful data management is one of the most critical aspects of plate-based high-throughput flow cytometry. Depending on the design of the plate, there may be as many as 96 separate FCS files. For the example used in this chapter, the FCS files require about 130 Mb of storage space per plate. The authors' data management strategy requires manual copying of files from the cytometer to a central hard drive that supports our core flow laboratory. From there, the files are copied one more time to a server in our corporate data center that provides a secure, version controlled location. We use Subversion and Eclipse along with standard backup processes implemented by our corporate data center to ensure the persistence and integrity of the original data.

Software

BD FACS™ CAP output from the cytometer is analyzed using semiautomated routines in R to ensure that the data are analyzed in an efficient, consistent, and objective fashion. The automated methods were developed to reflect the analysis that would be performed by an expert cytometrist using FlowJo. The results of the R-based analysis were validated by direct comparison to FlowJo results.

1. Use R 2.9.1 along with the R packages flowCore 1.10.0, plateCore 1.2.1, flowViz 1.8.0, KernSmooth 2.23-2, and ggplot2 0.8.3. flowCore, plateCore, and flowViz are flow-specific R packages developed to make processing, analyzing, and visualizing of flow data easier. In addition, plateCore provides a variety of built-in functions for managing plate-based flow cytometry data, which greatly increases the efficiency of analysis for BD FACS™ CAP. ggplot2 is an advanced graphics plot that greatly increases our ability to manage the many complex graphical displays required for data analysis.
2. In addition to these R packages, we use a collection of R-scripts and functions to manage the work flow and to automate much of the work that would have to be done by the cytometrist using standard GUI-driven software packages. All development of R-scripts is done using the integrated development environment (IDE) Eclipse along with the Eclipse plug-in, StatET. Standardized guidelines are followed for code development, and RUnit testing is performed to ensure correctness of the code. All scripts are securely maintained under version control using the technologies described earlier.

Compensation

Compensation can be done using either single-color staining of the same cells that are used for analysis, or BD CompBeads.

1. Conduct a preliminary study to identify the appropriate marker/dye combinations.
2. Run compensation samples in tubes each day and apply the same compensation results to the three plates that are run on that day. The current layout of the plate that is being used for new samples has wells devoted to compensation.
3. Compute the compensation matrix using plateCore and flow- Core. Results have been validated against FlowJo.

Morphology Gating

It is of critical importance to examine the health status of the samples used for the FACS™ CAP assay. Compromised viability of cells contributes to increased nonspecific binding of the antibodies and can confound interpretation of flow data. For special applications, a near-infrared LIVE/DEAD? fixable dead cell stain kit can be incorporated to evaluate the viability of the cells after the surface marker staining process. For most applications, however, it is sufficient to restrict data analysis to a cell population where dead cells, debris, and aggregates are excluded based on cell morphology.

Figure 1. A gating strategy is used to exclude dead cells, debris, and aggregates prior to data analysis.

1. Evaluate cell morphology based on measurements of FSC and SSC.
2. Select cells with normal morphology for further analysis using a standard gate. In general, cells are very stable across wells and multiple plates, and a single, fixed gate is used for one FACS™ CAP run.

Autofluorescence Correction

Autofluorescence is one of the major concerns in flow cytometry. One characteristic of cultured adherent cells is that they have a significant level of autofluorescence. Thus, an additional signal is present in each channel that cannot be corrected for by compensation and confounds any attempt to identify positive cells. Additionally, the output from R has been corrected by autofluorescence, which facilitates automated gating for cells with high autofluorescence.

Determining Gates Based on Isotype Controls

On each BD FACS™ CAP plate, there are a variety of controls, including isotype controls for selected immunoglobulin classes and subclasses. These isotype controls are used to set thresholds (negative control gates). Any cell that has a measured fluorescence intensity (FI) above the threshold is classified as "positive".

1. For each isotype control, set the initial threshold at the 99.5%- tile. Typically, the isotype controls are very similar from plate to plate for the same cell source and typically have CVs of less than 5%.
2. Review the representative density plots and make corrections for any thresholds that appear anomalous. Once the threshold/ gate has been finalized for each fluorescence channel, the percentage of positive cells reported reflects the fraction of cells with a fluorescent signal higher than that of 99.5% of the negative control cells.

Quality Control

The use of replicate plates of cells in the BD FACS™ CAP analysis primarily serves the purpose of quality control, as under normal circumstances, there is very little difference in the results among replicate plates. Although still relatively rare, the most common quality issue stems from fluidics problems that occur during the acquisition of cells by the flow cytometer.

1. To detect fluidics problems, examine empirical cumulative distribution function (ECDF) for either SSC or FSC.
2. Mark all the results corresponding to the well with QA problem as outliers and exclude them in any subsequent statistical analysis.

Summary Statistics

The primary summary statistic is the Percent Positive; that is, the percentage of cells that have a reading greater than the value of the negative control gate, as determined from the isotype controls. Depending on the particular application, a variety of other statistics may be of interest. Due to the highly reliable nature of flow cytometry, the variability in Percent Positive from plate to plate is usually quite small. For a typical donor in the present example, the mean standard error across three plates over all markers for Percent Positive was less than 1%. From a theoretical viewpoint, we would expect the Percent Positive to follow a binomial distribution when generated from independent samples of the same cells. The observed variability from plate to plate is generally lower than what would be expected if the plates represented true biological variability rather than simply technical replicates.

Normalization

For the purposes of understanding and visually comparing marker expression profiles, use the smoothed density. For the purposes of comparing multiple marker expression across multiple donors or multiple samples, combine the data from the replicate plates to get a single density.

1. When combining the data from multiple plates, adjust for minor differences in instrument settings or cell handling conditions by first calculating the average value of the negative control gates. Shift the values for each of the replicates by the difference between the control gate for that replicate and the average. Combine the data and calculate the final density.
2. When comparing data from multiple samples, adjust the densities in a similar fashion but do not recompute them. In this way, a single value can be used to represent the negative threshold gate across multiple samples, which greatly simplifies the display and makes it much easier to compare marker profiles across samples.

Categorization and Interpretation of Expression Levels

Because BD FACS™ CAP is a screening assay, the reported Percent Positive values should be interpreted with caution. In any case in which it is important to have an accurate estimate of the actual Percent Positive value for a given cell type, it is recommended that an optimized assay be used. Consequently, Percent Positive values are typically classified into one of four fairly broad categories; namely, negative, low, medium, or high.

(a) Negative corresponds to values of markers with Percent Positive £10%.
(b) The low category designates markers with 10% < Percent Positive £50%.
(c) The medium category designates markers with 50% < Percent Positive £85%.
(d) The high category designates markers with Percent Positive >85%.

Markers falling in the negative and high categories can be expected to stay in those categories when an optimized assay has been developed. The low and medium categories are expected to be less stable.

In many applications of flow cytometry in which a Percent Positive value is calculated, there are two distinct populations. In this case, one population is typically negative for the marker in question and one population is positive. The Percent Positive is the percent of cells falling in the positive population. In the case of adherent cells, it is more likely that there is only a single population, and the negative control gate falls in the interior of this population. When there is only a single population, a more accurate interpretation of the Percent Positive is "the percentage of cells with a measure intensity value greater than the negative control gate". Comparing Samples In the present example, BD FACS™ CAP was performed on human MSCs derived from three donors. Three replicate plates were assayed for each donor. As discussed earlier, the replicate plates represent technical (not biological) variability; the replication is primarily used for quality purposes. Data from replicate plates are combined before comparing results among the three donors. The test to compare the Percent Positive values among donors is based on the null hypothesis that the donors are not biologically different in regard to expression of a specific marker. In this case, the assumption is that the observed values for Percent Positive will follow a binomial distribution. The ratio of the observed variation in Percent Positive to the expected variation should follow a Chi-squared distribution. In this particular case, the distribution will only be approximate because the true Percent Positive is estimated from the data. Based on this chi-square distribution, a P-value is calculated for the test of the null hypothesis for each marker. Since there are more than 200 markers in the BD FACS™ CAP screen, the multiplicity problem must be addressed before computing final P-values. Multiplicity problem means that the risk of making false discoveries is extremely high when making many independent tests. To control for this false discovery rate, apply the method of Hochberg and Benjamini. After making this correction, an adjusted P-value of less than 0.05 will indicate that the samples (donors in this case) have different expression profiles.

1. When the samples are not statistically different, the markers across all samples were assigned a category (negative, low, medium, or high) using the average Percent Positive.
2. When the null hypothesis is rejected, assign the category "Different." 3. For any markers in which there are distinct subpopulations, mark these populations as "Heterogeneous" and do not consider the results of the statistical test to be meaningful because of the difficulty in interpreting the Percent Positive value.

Comparison of the Results to the Literature

Based on a review of the literature, the authors identified 12 markers that are commonly reported as being negative for bone marrow-derived MSCs and 15 markers that are commonly reported as being positive (see Note 4). Ten of the markers reported as negative in the literature are clearly negative for these donors. One marker (CD24) shows a low expression level and one marker (SSEA4) shows an intermediate level with differences among donors. Of the 15 markers reported as positive in the literature, 9 were approximately 100% positive, while 2 markers were negative. The other four markers were expressed at varying levels.

Heterogeneous Populations

A population is designated as heterogeneous in regard to expression of a particular marker if there are at least two distinct populations of cells. In this case, the density function will be bimodal, with each peak representing a different population of cells.

1. To identify multiple populations, use a peak-hunting algorithm on each density.
2. Carefully review markers with apparent multiple populations before officially classifying them as heterogeneous.
3. Results of a validation study suggest that subpopulations as small as 10% of the total population can be identified, as long as they are well separated. Smaller subpopulations may be identified if more cells are analyzed. For the MSC donors reported in the present example, no markers were identified as having heterogeneous expression.