Antigen-specific T cells may be detected and enumerated by short-term ex vivo antigen-specific stimulation followed by cytokine flow cytometry. Most frequently, intracellular IFN-γ is used to identify T cells specific for cytomegalovirus (CMV), Epstein-Barr virus or HIV. Some researchers use whole blood, others peripheral blood mononuclear cells (PBMC) in this assay; however, the performance of the two systems has never been directly compared. Blood was drawn from previously characterized healthy CMV-positive donors, and CMV-derived peptides or CMV lysate were used as stimulants. In an initial series of experiments, lithium-heparin was identified as the best coagulant to be used. Dose-response curves were established using concentrations between 0.1 and 40 μg/ml of peptides and between 0.1 and 20 μg/ml of virus lysate, respectively. IFN-γ-positive T cells were expressed as percent of the reference population, and frequencies measured in whole blood and PBMC were compared. Maximum responses were consistently higher in PBMC than in whole blood and were reached at lower concentrations of stimulant. In several instances, responses identified with PBMCwere not at all detected with whole blood. In summary, studies using whole blood in this type of assay are likely to underestimate the frequencies of antigen-specific T cells.
Despite the increasing availability of MHC/peptide tetramer reagents, the detection of intracellular cytokines remains a very important tool in antigen-specific T cell detection and analysis 1–7. Almost any stimulant can be chosen and, in contrast to tetramer staining alone 8, 9, it supplies some degree of "functional" information. In this respect, rapid cytokine induction evaluated by flow cytometry is similar to that evaluated by enzyme-linked immunospot (ELISPOT) assay; however, the latter does not reveal phenotypic features of responding cells, which at the time of read-out, are in fact absent 10, 11. In addition, for detecting intracellular cytokines, stimulation can be performed in whole blood and some researchers have used this simple and direct approach in basic research as well as in clinical studies 12.
However, we noticed that sometimes responses obtained with PBMC cannot be reproduced with whole blood, and sometimes donors classified as non-responders to CMV antigens by the whole-blood assay showed a response when PBMC were used instead. These observations prompted us to systematically compare the two approaches in a cohort of previously characterized CMV-responsive individuals. We chose CMV as a stimulant because CMV-specific T cell responses exist in a large proportion of the community. The choice of antigen as such, however, probably had little impact on the comparison of thetwo systems, since we essentially analyze and compare how efficiently a stimulating antigen is delivered to terminally differentiated effector cells 13 in whole blood and PBMC, or how conducive these environments are to their activation. It is, therefore, not primarily a study of the CMV-specific immune response. Intracellular IFN-γ was chosen as read-out, because it isa universal effector cytokine which is produced by both CD4 and CD8 T cells.
2.1 Donor selection based on previous analyses
Nine healthy blood donors with documented T cell reactivity to CMV lysate and/or known IE-1- and pp65-derived peptides had been selected in order to examine antigen-specific CD4 and CD8 T cell responses. The donors were also selected to provide a spectrum of responses ranging from less than 1% to several % of responder cells within the respective T cell population according to previous experiments.
2.2 Responses in CMV-seronegative individuals
Previously published results demonstrated that the T cell response to CMV lysate measured in PBMC correlates with seroreactivity, i.e. seronegative subjects show no CD4 T cell response to the lysate 1. We previously reported that the percentage of CD4 or CD8 T cells producing IFN-γ following stimulation with CMV-derived stimuli (lysate or peptides) in CMV seronegative individuals is usually very low (of the order of 0.02%) when using PBMC 5, 14. Suni et al. reported the respective frequencies to be similarly low in the whole-blood assay. Therefore, unspecific stimulating effects of CMV lysate or CMV-derived peptides in seronegative individuals do not seem to occur above the usual level of background stimulation observed in unstimulated samples from CMV-seropositive individuals (same order of magnitude) 5.
2.3 Choosing the best anticoagulant for performing dose-response curves
First, we examined the effect of sodium-heparin (Na-heparin), lithium-heparin (Li-heparin), and sodium-citrate (Na-citrate) on the whole-blood and the PBMC assay in three different individuals using 10 μg/ml of both CMV viral lysate and peptides (from previous studies these concentrations were known to cause maximum responses in most donors). Na-heparin is the anticoagulant used for this type of assay in several published studies 15, 16. Li-heparin is standard for routine blood testing in most hospitals in Germany and a number of European countries and provided in ready-made syringes, while Na-citrate is universally used for the analysis of blood clotting.
Fig. 1A shows the results of CD8 T cell activation with all three anticoagulants in a representative donor. The choice of anticoagulant clearly had no significant impact on the frequencies measured in the PBMC assay, which can probably be explained by the fact that the preparation procedure removes the anticoagulant. In whole blood higher responses were measured in two out of three donors when Li-heparin was used compared to Na-heparin (both CD8 and CD4 T cell responses); however, the differences were not striking. Fig. 1B shows results of CD4 T cell activation with virus lysate in whole blood in two donors, one of whom was the only donor with a slightly higher response with Na-heparin. Four additional experiments using 5 μg/ml of virus lysate in whole blood, however, confirmed that in a majority of experiments (six out of seven) the responses in Li-heparin-anticoagulated samples were somewhat higher than in Na-heparin-anticoagulated samples.
Because Li-heparin seemed to perform best with whole blood it was chosen as anticoagulant for the dose-response curves. By contrast, Na-citrate clearly suppressed reactivity in whole blood (Fig. 1A, B). With respect to Na-citrate it is important to note that in the PBMC assay calcium is replenished by the culture media (RPMI contains calcium) while it remains depleted in the whole-blood procedure.
2.4 Dose-response curves
Second, dose-response curves were established using 0.1, 1, 5 (in one case only), 10, and 20 μg/ml of CMV viral lysate and 0.1, 1, 10, and 40 μg/ml of the 9-mer and 15-mer peptides (Fig. 2A–C). As expected (and intended by the design of the study), the magnitudes of the T cell responses varied greatly between individuals. The marked differences between T cell frequencies measured in whole blood and PBMC, however, were highly consistent across the population of donors. Maximum responses in both CD4 and CD8 T cells were significantly higher and/or reached at significantly lower concentrations in PBMC than in whole blood. In some donors responses to certain peptides were below the detection limit in whole-blood samples while responses were readily identified in PBMC. These results therefore confirmed our previous observation that donors who had been classified as non-responders by the whole-blood assay were responders when the test was performed using PBMC.
The analysis of dose-response curves obtained with PBMC in previous experiments had indicated that concentrations of 1 μg/ml of a 9-amino acid peptide generally induced optimum or close-to-optimum responses 5; raising the concentration to 10 μg/ml did not have much of an effect. Series (a) and (b) in Fig. 2A show, however, that using 10 μg/ml instead of 1 μg/ml increased the response by approximately 100% and 50%, respectively. Also, series (f) in Fig. 2B shows an example of a donor who barely responded to the 15-amino acid peptide at 1 μg/ml, but had a strong response when 10 μg/ml was used. While we are satisfied from our previous data that for most donors 1 μg/ml is a close-to-optimum-working concentration for both 9- and 15-amino acid peptides in PBMC, these results show that there is a certain amount of variation between individuals.
Occasionally 15-amino acid peptides stimulate CD8 T cells much less than the corresponding 9-amino acid peptides; this becomes obvious by comparing series (b) in Fig. 2A with series (e) in Fig. 2B, which were obtained with the same donor. Such differences may depend on the positioning of the MHC-binding amino acid sequence within the 15-amino acid peptide and the efficiency of peptide clipping prior to external MHC class I loading 14. Examples (dot plots) of responses in whole blood and PBMC are shown in Fig. 3A, B for CD4 and CD8 T cells, respectively.
2.5 Variability of determinations
In order to determine an accurate coefficient of variation (CV) of this type of analysis it would have been necessary to run the assays in more than three replicates. A number of triplicate determinations were subject to relatively high (intra-assay) variation (e.g. Fig. 2C). Variation was more pronounced with virus lysate than with peptides and occurred both with PBMC and whole-blood stimulation. In a separate study the intra-assay variability of the PBMC approach was addressed systematically using ten parallel replicates. While for peptide stimulation the CV was around 12%, it was around 15% for virus lysate (manuscript in preparation); generally the CV was higher when the percentages of responding cells were smaller.
2.6 Statistical analysis
The overall performance of the two approaches was compared using the non-parametric Wilcoxon test for paired samples. This way the (median) frequency of IFN-γ-producing T cells obtained with each concentration of peptide or virus lysate in PBMC was directly compared to the corresponding (median) frequency obtained in whole blood in each donor (corresponding to a sample pair). The number of experiments in each series was too small to do this for each concentration individually. The difference between whole blood and PBMC was significant for 9-amino acid (p<0.05) and 15-amino acid peptides (p<0.01). It was not significant for virus lysate. The reason is that especially with virus lysate responses declined with high concentrations of antigen in some cases. For example, Fig. 2C, series (g–i) show that clearly more T cells were induced in PBMC with 1 μg/ml of virus lysate than with any concentration in whole blood; in contrast, more T cells were induced in whole blood than in PBMC when 20 μg/ml of virus lysate was used in series (g) or (h), and 10 and 20 μg/ml in series (i).
Soon after the development of a rapid flow cytometric approach to analyze antigen-specific T cells in PBMC, the use of whole blood for this procedure was reported 15, 16. One reason for using whole-blood cultures, apart from the reduction of work load compared to PBMC preparation, is the frequently encountered inference that whole blood provides a "more physiological" environment than cultures of PBMC in "artificial" media. The recognition, however, of antigen by peripheral T cells occurs generally in secondary lymphatic organs or in peripheral tissues. The environment provided by anticoagulated whole blood is similar to neither. One big difference between whole blood and RPMI media supplemented with FCS is the presence of granulocytes and platelets in blood, including the soluble factors secreted by these two cell types. Granulocytes may phagocytose material that can subsequently not be presented by APC. Moreover, the protein content and composition of whole blood and supplemented RPMI are different. It is possible that the presence of certain plasma proteins not (or at much lower concentration) present in PBMC preparations interferes with antigen uptake or direct MHC loading. Hydrophobic proteins or protein fragments may stick to cell membranes or proteins.
Whichever explanation is relevant, our results clearly show that higher doses of antigens are required to achieve the same degree of stimulation in whole blood compared to PBMC (measured in frequencies of activated T cells), and that in some experiments the same degree of stimulation was not reached even with much higher doses in whole blood. Depending on the stimulus used, this important difference makes the whole-blood assay not only more expensive but also less reliable. The fact that some responses obtained with PBMC could not be reproduced at all in whole blood is a clear illustration of the reduced sensitivity of the whole-blood assay, which may easily classify donors with small responses as "non-responders". It is interesting in this regard that with both whole blood and PBMC sometimes responses declined with high concentrations of antigen, especially with virus lysate (see Fig. 2C).
The fact that smaller doses of peptides are required to obtain "near-optimum" responses in PBMC implies that larger numbers of peptides can be efficiently tested than in whole blood. With peptides dissolved in DMSO (this is generally preferred for peptides at high concentrations), a toxic level of DMSO is reached with fewer peptides, if the concentrations have to be higher. This is particularly important for the use of protein-spanning peptide libraries 14. An additional disadvantage is that whole blood cannot be standardized or adjusted with respect to cell counts or any other components (it would not be whole blood any longer). If the white blood count is very low and the frequency of antigen-specific T cells small, the whole-blood approach is clearly not very promising unless large volumes are stimulated.
Interestingly, in 1998 Suni et al. 15 reported that slightly higher frequencies of cytokine-producing cells could be observed in whole blood than in PBMC cultures. However, for this study Suni et al. used autologous donor plasma as culture medium for PBMC (which is handy as it can be retrieved from the BD CPT preparation tubes). However, the same group later discovered that use of autologous plasma suppressed responses compared to standard culture media in this assay (Holden Maecker, personal communication). Therefore, these results cannot be compared to ours.
In order to explain the observed differences it may be proposed that the activation of T cells in whole blood and PBMC/RPMI cultures follow different kinetics and that the response in whole blood simply comes later. This is, however, not the case. The kinetics of this assay were carefully examined by Waldrop et al. in 1998 (PBMC) 17 and Namura et al. in 1999 (whole blood) 16. The optimum incubation times for stimulating IFN-γ responses in the whole-blood and PBMC assays were approximately 6–10 h 16, and 8–10 h 17, respectively.
Studies in PBMC have shown that the frequencies of antigen-specific T cells measured by intracellular IFN-γ staining come quite close to the frequencies measured by direct staining to the antigen-recognizing population using MHC class I/peptide tetramers 18. Importantly, IFN-γ- or other cytokine-producing cells are found only within the tetramer-positive population; however, a variable percentage of tetramer-positive T cells do not produce IFN-γ . From all available techniques, staining with tetramers is currently the most exact way of determining the frequencies of antigen-specific T cells, as it is independent of T cell activation 19, 20. Using PBMC in the above-discussed CFC approach therefore does not overestimate but in fact slightly underestimates the ‘real’ frequencies of antigen-specific T cells.
The reason for choosing IFN-γ as read-out in this study was that it is produced by effector T cells irrespectively of lineage. Choosing IL-2, by contrast, would have limited the study to CD4 T cells, since in our hands, CD8 T cells hardly ever respond with IL-2 when stimulated with CMV antigens. This is also true for IL-4 or IL-10. We would argue that in the system we used, IFN-γ production was the most expected outcome of triggering T cells, because this corresponds to the profile of terminally differentiated effector cells that is produced by an infection with CMV. What we measured was simply how effectively this outcome was produced. This is a function of whether the antigens we used were presented on the appropriate MHC molecules, whether T cells and APC interacted in the expected fashion, and, whether the culture conditions were appropriate. From this point of view, we would not expect to find essentially different results in a system that was for example more conducive to measuring intracellular IL-4. Nevertheless, this study may prompt others to examine the system they use in a similar way as we did.
In summary, this study indicates that the use of PBMC in this rapid ex vivo cytokine induction assay is preferable over using whole blood in a variety of situations. Clearly, if the response to an antigen is known in a particular individual, the whole-blood approach may be suitable for monitoring this response. However, the likelihood that screening studies performed with whole blood will underestimate antigen-specific T cell frequencies or even miss responses altogether is high. Especially classifications of donors/patients into responders or non-responders based on the whole-blood approach ought to be revised. It is not excluded that certain changes in the experimental set-up will be able to improve the performance of the whole-blood assay in the future. It is important, therefore, to observe that this is a comparison of two standard assays the way they are being used in the scientific community. It is not an absolute statement against the use of whole blood.
4 Materials and methods
In order to examine the effect of different anticoagulants on the assays, peripheral venous blood from five healthy CMV-seropositive (IgG) donors was drawn into 5.5-ml or 10-ml containers (Monovette; Sarstedt, Nümbrecht, Germany) containing either 0.1 ml Na-citrate or 15 IU of Li-heparin per ml of blood. For testing Na-heparin as anticoagulant, 75 IU of Na-heparin (Liquemin; Roche, Germany) was filled into 10-ml neutral containers (Monovette, Sarstedt) prior to drawing blood.
4.2 Peptides and viral lysate
Purified CMV viral lysate (ABI, Columbia, MD) was diluted to 0.25 mg/ml with sterile phosphate-buffered saline (PBS, Dulbecco's) and frozen in aliquots at –70°C. Antigenic peptides of IE-1 (Swiss-Prot P13202) and pp65 (Swiss-Prot P06725) of 9-amino acid and 15-amino acid length were produced in-house (standard Fmoc synthesis). Peptide stocks (80 mg/ml) were produced by dissolving freeze-dried peptides in DMSO (Pierce, Rockford, IL) and kept at –70°C. Working solution was prepared from stock solution by dilution with DMSO/PBS.
4.3 Antibodies and staining
The following monoclonal antibodies were obtained from BD Biosciences (San Jose, CA): fluorescein isothiocyanate-conjugated anti-IFN-γ, phycoerythrin (PE)-conjugated anti-CD69, peridinin chlorophyll protein (PerCP)-conjugated anti-CD3, and allophycocyanin-conjugated anti-CD8. PE-conjugated anti-CD4 was purchased from Immunotech (Coulter, Marseille, France).
4.4 Cell preparation
Blood was obtained from healthy volunteers following informed consent and divided in two aliquots each. For the whole-blood assay, aliquots of 1 ml anticoagulated whole blood were placed in closed 14-ml Cellstar polypropylene tubes (Greiner, Frickenhausen, Germany) and kept at room temperature (RT), while PBMC were prepared from the remaining specimens.
4.5 Whole-blood procedure
Peptides or viral lysates were added to the aliquots in volumes of 1, 10, 20, and 40 μl (working dilutions of 1 mg/ml and 0.1 mg/ml). The final concentrations of stimulants were 0.1, 1, 5 (in one case), 10, and 40 μg/ml for peptides and 0.1, 1, 10, and 20 μg/ml for viral lysate. DMSO concentrations in all assays including unstimulated controls were kept below 0.1% (vol/vol). All samples were placed at a 5°-slant in a standard incubator (37°C, humidified 5% CO2 atmosphere). After 2 h Brefeldin A (BFA; Sigma, Munich, Germany) was added (2 μl of a 5 mg/ml stock solution dissolved in DMSO). The final concentration of BFA was 10 μg/ml. After additional 4 h the cells were washed with ice-cold PBS (430×g, 8 min, 4°C) and vortexed for 30 s. FACSLysing Solution (BD) was added according to the manufacturer's instructions and tubes were kept for 10 min at RT. Cells were then spun down, the supernatant was removed by decanting, and the pellet was resuspended carefully. Next, 4 ml of washing buffer [PBS containing 0.5% (w/vol) bovine serum albumin (Serva, Heidelberg, Germany) and 0.1% (w/vol) sodium azide (Serva)] was added and the cells were spun down (430×g, 8 min, 4°C); the supernatant was removed by decanting and the tubes were blotted dry on a paper towel.
Permeabilization was performed using permeabilizing buffer consisting of double-concentrated FACSLysing Solution (BD) containing 0.05% (vol/vol) Tween 20 (Sigma, Steinheim, Germany). This combination has previously been found to be as effective as commercial BD permeabilizing solution for permeabilization; however, it could also be used as a time-saving one-step fixation-permeabilization reagent for PBMC (see below). For this purpose 1 ml of permeabilizing buffer was added to the cells and all tubes were incubated for 10 min at RT in the dark. Following an additional wash step, simultaneous surface and intracellular antibody staining was performed for 30 min on melting ice in the dark. All samples were processed in triplicate.
4.6 PBMC assay
PBMC were prepared by standard Ficoll-Paque (Pharmacia, Uppsala, Sweden) gradient centrifugation. Cells were then washed three times with PBS and resuspended in RPMI 1640 medium (Biochrome, Berlin, Germany) containing 2 mmol/l L-glutamine (Biochrome), 10% (vol/vol) heat-inactivated FCS (Biochrome), and 100 IU penicillin/streptomycin ("supplemented media") at a concentration of 2.0×106cells/ml. Peptides or viral lysates were added in volumes of 1, 10, 20, and 40 μl (working dilutions of 1 mg/ml and 0.1 mg/ml) to 100 μl supplemented media placed in Falcon 2052 tubes (BD) prior to adding 400 μl of the cell suspension. The final concentrations of stimulants were 0.1, 1, 10, and 40 μg/ml for peptides and 0.1, 1, 5 (in one case), 10, and 20 μg/ml for viral lysate. DMSO concentrations in all assays including unstimulated controls were kept below 0.1% (vol/vol).
Samples were then placed in a standard incubator at a 5°-slant (37°C, humidified 5% CO2 atmosphere). After 2 h, 2 μl of a 5 mg/ml stock solution of BFA (Sigma) in DMSO was dissolved in 500 μl of warm supplemented medium and added to the culture (the final concentration of BFA was 10 μg/ml). After additional 4 h, the cells were washed with 3 ml ice-cold PBS (430×g, 8 min, 4°C) and resuspended in 3 ml of PBS containing 2 mM EDTA (Sigma), incubated for 10 min at 37°C (water bath), spun down (430×g, 8 min, 4°C), and vortexed for 30 s. Cells were then washed with washing buffer, the supernatant was removed by decanting, and tubes were blotted dry on a paper towel. Fixation/permeabilization was performed by adding 1 ml of permeabilizing buffer (see Sect. 4.5) to the cells prior to vortexing. Antibody staining followed exactly the same protocol as described for the whole-blood assay.
4.7 Flow cytometric analysis
Data acquisition was carried out on a Becton Dickinson FACSCalibur flow cytometer using CellQuest and Paint-A-Gate software. Lymphocytes (250,000; live gate) were acquired for each sample. For data analysis, IFN-γ+/CD3+/ CD4+ or IFN-γ+/CD3+/CD8+ events were gated and percentages of the total CD4+ and CD8+ T cells were determined. Percentages of IFN-γ+ events in corresponding gates from unstimulated control samples were subtracted. Isotype controls were not performed, since previous results in >1,000 samples showed that isotype-stained samples were indistinguishable from unstimulated IFN-γ-stained samples (manuscript in preparation); however, they markedly increased cost and labor. The analysis strategy (gating) is explained in Fig. 4.