Standardization of cytokine flow cytometry assays

Abstract

Background: Cytokine flow cytometry (CFC) or intracellular cytokine staining (ICS) can quantitate antigen-specific T cell responses in settings such as experimental vaccination. Standardization of ICS among laboratories performing vaccine studies would provide a common platform by which to compare the immunogenicity of different vaccine candidates across multiple international organizations conducting clinical trials. As such, a study was carried out among several laboratories involved in HIV clinical trials, to define the inter-lab precision of ICS using various sample types, and using a common protocol for each experiment (see additional files online).

Results: Three sample types (activated, fixed, and frozen whole blood; fresh whole blood; and cryopreserved PBMC) were shipped to various sites, where ICS assays using cytomegalovirus (CMV) pp65 peptide mix or control antigens were performed in parallel in 96-well plates. For one experiment, antigens and antibody cocktails were lyophilised into 96-well plates to simplify and standardize the assay setup. Results ((CD4+)cytokine+ cells and (CD8+)cytokine+ cells) were determined by each site. Raw data were also sent to a central site for batch analysis with a dynamic gating template. Mean inter-laboratory coefficient of variation (C.V.) ranged from 17-44% depending upon the sample type and analysis method. Cryopreserved peripheral blood mononuclear cells (PBMC) yielded lower inter-lab C.V.'s than whole blood. Centralized analysis (using a dynamic gating template) reduced the inter-lab C.V. by 5-20%, depending upon the experiment. The inter-lab C.V. was lowest (18-24%) for samples with a mean of > 0.5% IFNgamma + T cells, and highest (57-82%) for samples with a mean of < 0.1% IFNgamma + cells.

Conclusion: ICS assays can be performed by multiple laboratories using a common protocol with good inter-laboratory precision, which improves as the frequency of responding cells increases. Cryopreserved PBMC may yield slightly more consistent results than shipped whole blood. Analysis, particularly gating, is a significant source of variability, and can be reduced by centralized analysis and/or use of a standardized dynamic gating template. Use of pre-aliquoted lyophilized reagents for stimulation and staining can provide further standardization to these assays.

Figure 1

Experimental design. (A) Schematic of protocol for Experiments 1–3, performed using liquid antigens and antibodies. (B) Schematic of protocol for Experiment 4, performed using lyophilised antigen and antibody plates.

Figure 2

Manual versus automated gating templates. (A) Representative manual analysis of a CEF-stimulated sample from Experiment 4. Sequential gates on small lymphocytes, CD3⁺ cells, and CD3⁺CD8⁺. cells are applied and the percent CD69⁺IFNg⁺ cells are determined from a plot gated on all of these regions. (B) Dynamic gating template for the same data file as above. Sequential dynamic gates ("Snap-To" gates) are applied as above, except that negative populations are also gated so as to provide a boundary for the movement of the positive region. The percent CD69⁺IFNg⁺ cells obtained is very similar to that obtained by manual gating in this example, since manual gating was performed so as to include CD3^dim and CD8^dim cells.

Figure 3

Results of Experiment 1 (fixed activated whole blood). IFNγ-positive cells in response to SEB or CMV pp65 peptide mix are expressed as a percentage of CD4⁺ or CD8⁺ T cells. Results from each site are indicated as a circle, with median responses for each sample (105, 950, and 1040) indicated by a horizontal bar. The C.V. for each sample is listed across the top of each panel, along with the mean C.V. for that set of samples.

Figure 4

Results of Experiment 2 (shipped whole blood). IFNγ-positive cells in response to SEB or CMV pp65 peptide mix are expressed as a percentage of CD4⁺ or CD8⁺ T cells. Results from each site are indicated as a circle, with median responses for each sample (105, 1040, and 1090) indicated by a horizontal bar. The C.V. for each sample is listed across the top of each panel, along with the mean C.V. for that set of samples.

Figure 5

Viabilities and recoveries for cryopreserved PBMC samples used in Experiment 3. Error bars represent SEM of the 6 sites participating.

Figure 6

Results of Experiment 3 (cryopreserved PBMC). IFNγ-positive cells in response to SEB or CMV pp65 peptide mix are expressed as a percentage of CD4⁺ or CD8⁺ T cells. Results from each site are indicated as a circle, with median responses for each sample indicated by a horizontal bar. Sample names are listed along the X axis. The C.V. for each sample is listed across the top of each panel, along with the mean C.V. for that set of samples.

Figure 7

Results of Experiment 4 (cryopreserved PBMC with lyophilized reagents). Cytokine-positive cells in response to CEF peptides or CMV pp65 peptide mix are expressed as a percentage of CD4⁺ or CD8⁺ T cells. Results from each site are indicated as a circle, with median responses for each sample indicated as a horizontal bar. Sample names are listed along the X axis. The C.V. for each positive sample is listed across the top of each panel, along with the mean C.V. for that set of samples. (A) Data reported by individual sites. (B) Data after centralized analysis using a dynamic gating template. Cocktail 1 consisted of CD4 FITC/IFNγ +IL-2 PE/CD8 PerCP-Cy5.5/CD3 APC. Cocktail 2 consisted of IFNγ FITC/CD69 PE/CD4 PerCP-Cy5.5/CD3 APC, and cocktail 3 consisted of IFNγ FITC/CD69 PE/CD8 PerCP-Cy5.5/CD3 APC. (C) Control cell data from Experiment 4. Activated, processed, and stained PBMC were lyophilised and run as controls by each site in Experiment 4. These cells had been stained with cocktail 1, 2, or 3. The left panel shows data reported by individual sites; the right panel, data after centralized analysis using a dynamic gating template. Most of the inter-lab variability was removed by this method of analysis.

Figure 8

Background cytokine-producing cells by experiment. Cytokine-positive cells in the absence of stimulus are expressed as a percentage of CD4⁺ or CD8⁺ T cells. Each symbol represents the background from a single sample processed by a single site. Medians are shown by a horizontal bar. Data from experiment 4 are for cocktails 2 and 3 only, to be comparable with experiments 1–3.

Figure 9

Comparison of liquid and lyophilised reagents. Comparative results are shown with backgrounds subtracted; no significant differences in backgrounds were seen with liquid versus lyophilised reagents (data not shown). (A) Data from one site that compared cocktail 1 (CD4 FITC/IFNγ +IL-2 PE/CD8 PerCP-Cy5.5/CD3 APC) in liquid and lyophilised form in Experiment 4. Black bars indicate liquid antibodies, grey bars indicate lyophilised antibodies. Error bars indicate SEM of duplicate wells. (B) Combined comparison of liquid antigen + liquid antibodies versus lyophilised antigen + lyophilised antibodies. Whole blood was activated with either SEB or pp65 peptide mix, and the percentage of IFNγ⁺ cells (CD4⁺ or CD8⁺) were compared with liquid versus lyophilised reagents (left panel). A similar comparison was made for the mean fluorescence intensity (MFI) of IFNγ⁺ cells (CD4⁺ or CD8⁺) (right panel). Similar results were obtained using PBMC (not shown).

Figure 10

S.D. versus mean for all assays. A linear relationship between S.D. and mean is expected based upon counting statistics [23]. This expected relationship (for a data set of 40,000 events) is shown by the solid black line. The actual data from the four experiments is shown in the symbols. Data from experiment 4 are for cocktails 2 and 3 only, to be comparable with experiments 1–3. The difference (in the Y dimension) between the data points and the solid line represents variability from sources other than counting statistics. Note that the data points for all three assay types cluster together, indicating that variability is similar for all three assay types. When individual analysis variability is removed (B), there is a slight tendency toward lower variability with cryopreserved PBMC (solid circles), and higher variability with shipped whole blood (open squares). The tendency toward lower variability is more pronounced in the experiment using cryopreserved PBMC with lyophilised reagents (panel B, open circles).