Major histocompatibility complex (MHC)-peptide tetrameric complexes (tetramers) are valuable tools for detecting and characterizing peptide-specific T cells. Because the frequency of these cells is generally very low, it may be difficult to discriminate between nonspecific and specific tetramer binding.
A four-color flow cytometric assay that simultaneously measures tetramer, CD3, CD8, and CD14 was used to investigate the sensitivity and specificity of MHC class I tetramer staining. This was accomplished by using the influenza virus matrix protein peptide, GILGFVFTL (FLU), as a model recall antigen and the human immunodeficiency virus (HIV) reverse transcriptase peptide, ILKEPVHGV (HIV), as a model novel antigen. Peripheral blood mononuclear cells (PBMC) from 31 HLA-A2.1+ and 10 HLA-A2.1− healthy individuals were stained with the tetramers.
The lower limit of detection was established at approximately 1/8,000. In HLA-A2+ PMBC, frequencies of tetramer-positive CD8+ T cells were log normally distributed and were high for FLU (1/910) but low for HIV (1/6,067). A novel competition assay, in which tetramer binding was shown to diminish subsequent staining with anti-CD3 antibody, was used to confirm the specificity of tetramer binding to the T-cell receptor (TCR) complex. The competition assay was validated by evaluating several anti-CD3 antibodies and showing that in PBMC from HLA-A2− subjects, spurious tetramer-positive events (1/20,000) failed to compete with CD3 binding. For the “recall” FLU tetramer, the degree of competition was proportional to the frequency, suggesting a selection of high avidity cells. Although CD3 competition was also highly correlated with the intensity of tetramer staining, competition allowed the identification of false positive cases with relatively high tetramer staining intensity.
Soluble peptide-major histocompatibility complexes (MHC) have a fast dissociation rate from the T-cell receptor (TCR). Multimeric peptide-MHC complexes bind to more than one TCR on a specific cell and, therefore, have slower dissociation rates (1, 2). Multimeric peptide-MHC complexes (generically referred to as tetramers because of their 4:1 molar ratio of biotinylated protein complexes to streptavidin) are proving invaluable as fluorescent reagents for enumeration, characterization, and isolation of peptide-specific T cells. They have afforded many advantages over previous techniques, particularly the ability to directly quantify and phenotype antigen-specific T cells with minimal in vitro manipulation (3–11). However, few studies have investigated physical characteristics that underlie tetramer binding. Wheelan et al. (12) described temperature as a variable affecting the specificity and intensity of tetramer staining. Bodinier et al. (13) determined that nonspecific tetramer binding can occur through interaction with the CD8 coreceptor. In addition, Yee et al. (14) correlated tetramer staining intensity with T-cell avidity and demonstrated that tetramers can be used to selectively identify high avidity cytotoxic T lymphocytes (CTL). The majority of investigators have used relatively simple two- or three-color staining strategies to identify CD8+ tetramer-positive events. Lee et al. (15) used multiparameter flow cytometry to hone the specificity of tetramer staining by eliminating CD4+, CD19+, and CD14+ cells from their analysis. However, this approach has not been widely adopted, owing perhaps to the complexity of their 10-color flow assay. Multiparameter analysis may not be necessary when the goal is to characterize the tetramer binding properties of T-cell lines or clones or to evaluate the in vivo response to strong antigens such as viral peptides. However, when the objective is to evaluate the in vivo frequency of T cells to novel antigens, self-antigens, or tumor antigens (14–22), for which tetramer binding frequencies are low and staining patterns may be ambiguous, it is important to establish a method that minimizes background staining and allows tetramer binding specificity to be evaluated as an independent parameter.
In this study, we addressed these issues using MHC class I tetrameric complexes containing a human immunodeficiency virus (HIV) peptide and an influenza virus (FLU) peptide as models for novel and recall antigens, respectively. The aim was to optimize analysis for future studies with expected low frequencies of peptide-specific T cells. A four-color flow cytometry-based technique is described, allowing the reduction of background staining and the confirmation of the specificity of tetramer binding to the TCR.
MATERIALS AND METHODS
Peripheral Blood Mononuclear Cells
Peripheral blood or a leukapheresis product was obtained from 31 HLA-A2.1+ and 10 HLA-A2.1− normal healthy donors. Peripheral blood mononuclear cells (PBMC) were isolated by centrifugation over Ficoll-Hypaque gradients (Amersham Pharmacia Biotech, Piscataway, NJ). Leukapheresis products were prepared by the Institute of Transfusion Medicine (Pittsburgh, PA). The study was approved by the institutional review board (IRB) at the University of Pittsburgh and written informed consent was obtained from each individual. All products were tested and found negative for HIV-I antigens and antibodies to HIV.
PBMC were frozen at a concentration of 50 × 106 cells/mL in freezing medium consisting of human AB serum (Pel-Freeze, Brown Deer, WI) plus 10% dimethylsulfoxide (DMSO; Fisher Scientific, Pittsburgh, PA). Cells were held overnight at -80°C and transferred to liquid N2. Cells were rapidly thawed at 37°C and an excess of cold Aim-V medium (Life Technologies, Grand Island, NY) supplemented with 20% human AB serum was added to the cells. The cells were washed twice, adjusted to 5 × 106/mL, and maintained for 2–4 h at room temperature. Clumps were carefully removed and viability of the remaining cells was measured by the trypan blue dye exclusion test. Average viability was 70–80%.
PBMC were phenotyped for expression of HLA-A2 molecules by flow cytometry, using the anti-HLA-A2 monoclonal antibody (mAb), BB7.2 (ATCC, Manassas, VA), and an IgG isotype as a control. The verification of the A2.1 subtype was performed in the laboratory of Dr. Adriana Zeevi using polymerase chain reaction with sequence-specific primers (PCR-SSP) as previously described (23). Briefly, DNA was obtained from PBMC using sodium dodecyl sulfate (SDS) and proteinase K. After removal of protein contaminants with a saturated salt solution, DNA was precipitated using 2 volumes of ethanol. Following washing and drying, the DNA pellet was reconstituted and quantified by reading the optical density (OD) at 260 nm. The extracted DNA was subjected to PCR-SSP (Dynal, Oslo, Norway) with primers that distinguish the specific allelic polymorphisms. Each primer set was designed to give an amplified fragment of a specific size, which was detected by gel electrophoresis and ethidium bromide staining. Patterns of positive and negative amplification yield the relevant genotype. A total of 31 HLA-A2.1+ and 10 HLA-A2− healthy individuals were identified and included in the study.
Tetrameric Peptide-MHC Class I Complex (Tetramer) Assay
Tetramers were obtained through the National Institute of Allergy and Infectious Diseases (NIAID) Tetramer Facility and the National Institutes of Health (NIH) AIDS Research and Reference Reagent Program; stock solutions contained 0.5 μg monomers per milliliter. Tetramer preparation was described in detail by Altman et al. (1). Peptides provided to the NIAID were either GILGFVFTL, an influenza matrix immunodominant peptide (residues 58–66), or the HIV-1 reverse transcriptase peptide (pol 476–484), ILKEPVHGV.
To minimize background staining, each tetramer was titered and used at the lowest concentration that still gave a clearly discernible positive population in a donor vaccinated for influenza (FLU 58-66 tetramer) and in an HIV-infected individual (pol 476–484; 10). The final dilution of both preparations during staining, relative to the stock reagent supplied by the NIH, was 1/300. Within a twofold range of tetramer concentrations bracketing the concentrations used here, the frequency of tetramer-positive events and competition of CD3 binding were stable and tetramer fluorescence intensity was within 80% of that obtained at saturating concentrations. At saturating concentrations, CD3 competition decreased, fluorescence intensity of tetramer-positive cells increased, and tetramer frequency increased, the latter due chiefly to an increased number of tetramer-dim events.
The default panels for these studies were CD14-FITC (RMO52; Immunotech, Miami, FL), tetramer-PE, CD3-ECD (HIT3a; Beckman Coulter, Miami, FL), and CD8-PC5 [SFCI21Thy2D3(T8), Beckman Coulter]. Additionally, the following antibodies were used: anti-human α/β FITC (WT31; Becton Dickinson, Mountain View, CA); anti-human CD2 FITC (Leu-5b; Becton Dickinson); anti-human CD3 FITC (HIT3a; PharMingen, San Diego, CA); anti-human CD3 FITC (UCHT1; Immunotech); anti-human CD3 FITC (SP34; PharMingen); anti-human CD3 FITC (SK7; Becton Dickinson); anti-human CD5-FITC (BL1a; Immunotech); anti-human CD14-ECD (RMO52; Coulter); anti-human CD45RA-PC5/TC (MEM56; Caltag, Burlingame, CA); and anti–CD45RO-FITC (UCHL1, Caltag).
Flow Cytometry Analysis
Immediately prior to staining, cells were washed twice with the staining medium consisting of phosphate-buffered saline (PBS) + 0.1% (w/v) bovine serum albumin (BSA) + 0.1% (w/v) sodium-azide and resuspended at a concentration of 5 × 106/mL in a volume of 150 μl. Tetramer (5 μl of 1:10 dilution of stock solution) was added at room temperature for 30 min, followed by a 30-min incubation with antibodies (7.5 μL of each) at 4°C. After two additional washes, the cells were resuspended in approximately 1 mL of 0.5% methanol-free formaldehyde in PBS. At least 1 × 106 events were collected using a four-color Coulter Epics XL cytometer set on low or medium flow rate at a maximum of 1,000 events per second. Flow cytometry data were analyzed in real time using Beckman-Coulter System II software. In initial experiments, the region defining tetramer-positive events was determined by evaluating PBMC stained with the antibody panel, but without tetramer. This region was held constant throughout the analysis. Data were saved as FCS 2.0 Listmode files for subsequent reanalysis in System II or WinList (Verity Software House, Topsham, ME).
Tetramer-positive cells were quantified by flow cytometry and expressed as frequencies (e.g., 1/1,000) or reciprocal frequencies (e.g., 1,000). We examined raw reciprocal frequency data and log-transformed reciprocal frequency data using normal probability plots (24). For both tetramers, the log-transformed data were better modeled by the normal distribution. Accordingly, descriptive statistics (means, SD, confidence intervals [CI]) and statistical analyses (Student's t-test, two-tailed) were performed on log-transformed reciprocal frequencies. For tetramer frequencies as well as competition of antibody binding, lower limits of detection were established by estimating the 99th percentile (geometric mean plus 2.6 SD) of responses measured in PBMC from HLA-A2- subjects. Linear regression was performed by the methods of least squares. In regression plots, the line of best fit and the 95% CI about the line of best fit are shown. The coefficient of correlation (r2), the slope of the regression line, and the P value associated with the slope are reported as regression summary statistics. Statistical analysis and statistical graphics were performed using Systat Version 9 (SPSS, Chicago, IL).
Elimination of Nonspecific Tetramer Binding Events
To increase the specificity of tetramer binding, we developed a gating strategy to eliminate monocytes and apoptotic and necrotic cells, all of which can potentially bind tetramer or antibody nonspecifically. PBMC from a healthy HLA-A2.1+ donor were incubated with the FLU tetramer (PE), followed by mAbs against CD14 (FITC), CD3 (ECD), and CD8 (PC5). Figure 1 shows that sources of nonspecific binding (e.g., CD14+ monocytes in Fig. 1A, dead cells in Fig. 1B, or other CD3− cells in Fig. 1C) were eliminated by compound gating. Cells with low tetramer staining intensity were also eliminated. A clearly discernible cluster of tetramer-positive/CD8+ T cells emerged with a mean fluorescence intensity (MFI), which was at least one log above the MFI of the negative population. Compound gating also permitted us to use a liberal lymphocyte scatter gate (Fig. 1B) that would not exclude large activated T cells. Without these compound gates, a “smear” of tetramer-positive CD8+ cells is often detected, including a significant number of cells with low tetramer staining intensity. For these reasons, the described gates were used for all subsequent experiments performed with the panel of antibodies including anti-CD14, anti-CD3, and anti-CD8.
Log Transformation of the Data and the Use of the Geometric Mean
It is essential to understand the distribution of data before descriptive statistics can be generated and statistical tests can be applied. Figure 2 shows a histogram of the raw and log-transformed HIV frequency data. Superimposed over the data are the normal distributions inferred from the sample mean and SD. The raw data are left skewed and does not conform to a normal distribution, invalidating the use of parametric statistics. Log transformation resulted in a more symmetric data distribution, with the mode and median values closer to the mean. For the two peptides used in this study and for others (data not shown), we have found that taking the log of the reciprocal frequency is optimal for bringing the data into conformation with a normal distribution.
Definition of the Lower Limit of Detection of Tetramer-Positive T Cells
In order to establish the lower detection limit for tetramer binding in HLA-A2.1+ individuals, we stained PBMC obtained from 10 HLA-A2− individuals. Tetramer binding to peptides representing a recall antigen (FLU) and a novel antigen (HIV) was examined. Despite the application of the gating strategy described above, low levels of FLU and HIV tetramer-positive CD3+ CD8+ T cells were detected in PBMC of HLA-A2− donors (geometric means = 1/22,131 and 1/23,551, respectively). Because these events were nonspecific by definition, we established a cutoff for the lower detection limit of this assay at the upper 99th percentile of tetramer-positive CD8+ T cells in HLA-A2− individuals. Using this cutoff, the lower limit of detection was 1/7,805 (pooled data for FLU and HIV tetramers). This cutoff was applied to all data obtained from testing the 10 HLA-A2− and 31 HLA-A2.1+ individuals (Fig. 3). None of the HLA-A2− individuals had frequencies of FLU or HIV tetramer-positive CD8+ T cells exceeding the cutoff. However, all HLA-A2.1+ individuals had detectable frequencies of FLU tetramer-positive CD8+ T cells (mean = 1/910) and 23 of 31 had detectable frequencies of HIV tetramer-positive CD8+ T cells (mean = 1/6,067).
CD45 Isoform Expression
In order to discriminate between putative naive or memory subsets in tetramer-positive T cells, we determined CD45 isoform expression in the PBMC of five subjects. Cells were stained with either FLU or HIV tetramer followed by CD45RO-FITC, CD8-ECD, and CD45RA-PC5. Monocytes and high side scatter NK cells were eliminated from the analysis by using a tight lymphocyte light scatter gate. For the recall antigen, FLU, the majority of CD8+ tetramer-positive cells were CD45RObright/CD45RA− (76.7 ± 2.7%). For the novel antigen, HIV, CD8+ tetramer-positive cells were predominantly CD45RO−/CD45RAbright (71.9 ± 9.4%).
Competition Between Tetramer and Anti-CD3 Antibody Binding
During analysis of the data set, we consistently noted that in PBMC from HLA-A2.1+ subjects, tetramer-positive events stained dimmer for CD3 than did tetramer-negative events (Fig. 1C). This was not the case for the spurious tetramer-positive events seen in the PBMC from HLA-A2− subjects (see below). This observation suggested that tetramer binding to the TCR interfered with binding of anti-CD3 antibody to CD3. To determine whether this phenomenon could be exploited as a marker for specificity of tetramer binding to the TCR complex (as opposed to nonspecific binding to some other surface structure), we quantified the extent to which tetramer binding competed with CD3 binding on CD8+ T cells according to the formula:
where CD3Mnl is the log MFI of CD3 staining of the population designated by the subscript.
To confirm and further characterize competition between tetramer and anti-CD3 antibody binding, we performed a series of experiments in which the anti-CD3 antibody (clone HIT3a) was exchanged for a variety of other antibodies directed to the epsilon chain of the TCR, TCR α/β, CD2, or CD5. Results from testing PBMC obtained from three HLA-A2.1+ donors indicated that binding of anti-CD3 clones HIT3a, UCHT 1, and SP34 was comparably inhibited by the FLU tetramer (Table 1). No consistent effect was observed with the anti-CD3 mAb SK7, anti-CD2, or anti-CD5. Interestingly, binding of antibodies to TCR α/β was consistently enhanced. The data indicate that tetramer binding to the TCR results in steric changes that influence binding of antibodies directed at some, but not all, epitopes of CD3 epsilon.
Table 1. Competition of Anti–T-Cell Antibodies With FLU Tetramer in PBMC From HLA-A2.1+ Donors†
Percent competition in CD8+ T cells (n = 3 HLA-A2+ normal donors)
Cells were stained with FLU tetramer-PE followed by mAbs. Results are means ± SD of results from three individuals. Negative competition, reported for CD2 and TCR α/β, indicates that anti–T-cell mAb staining was brighter in tetramer-positive events. Student's t-test (two-tailed) was used to determine whether competition was significantly different than 0%:
P < 0.025,
P < 0.01,
P < 0.005.
In addition to the designated anti–T-cell mAb, which was FITC conjugated, cells were stained with anti–CD14-ECD and anti–CD8-PC5.
Optimal competition of anti-CD3 binding was dependent on the staining sequence. When HLA-A2.1+ PBMC were incubated with the FLU tetramer first, followed by the incubation with the anti-CD3 antibody, competition was significantly greater (P = 0.008, paired t-test) than that obtained with the inverse approach (first antibodies, then tetramer; Table 2). Accordingly, CD3 staining intensity was significantly lower (P = 0.007) and tetramer staining was significantly brighter (P = 0.0004). In this small series, the estimated frequency of FLU tetramer-positive CD8+ T cells was also somewhat lower when cells were stained first with anti-CD3. However, the difference was not statistically significant (P = 0.06). These findings further support the conclusion that the diminution of CD3 binding is due to competition by the tetramer.
Table 2. Optimal Competition of Anti-CD3 Binding by Tetramer Requires That Tetramer Binding Precede Staining With Anti-CD3 (HIT3a-ECD)†
Order of reagent addition
Results are means of four independent experiments. Student's paired t-test (two-tailed) was used to determine whether competition was significantly different when the order of staining was changed:
Because there was no detectable competition between the tetramer and anti-CD3 antibody binding for CD8+ T cells in 9 of 10 HLA-A2.1− individuals (Fig. 4), we were able to define a cutoff based on the 99th percentile of CD3 competition in these HLA-A2− subjects (3.2%). CD3 competition in excess of this cutoff was considered significant. For the FLU tetramer, 28 of 30 subjects displayed significant CD3 competition. For the novel HIV peptide, there was significant competition of CD3 binding in only 11 of 31 subjects. Competition was strongest on CD8+ FLU tetramer-positive T cells (mean = 26.5%) and was lower for CD8+ HIV tetramer-positive T cells (mean = 4.7%).
When reciprocal frequencies of tetramer-positive CD8+ T cells were plotted against tetramer competition of anti-CD3 binding, a strong correlation was noted for FLU tetramer-positive CD8+ T cells (r2 = 0.425, slope = -34.7, P = 0.0004, Fig. 5, left panel). This relationship was not seen for the HIV tetramer (Fig. 5, right panel). This observation is consistent with the interpretation that T-cell expansion driven by repeated in vivo exposure to antigen results in selection of T cells that bind the tetramer more avidly and thus compete anti-CD3 binding more effectively.
In keeping with this interpretation, CD3 competition was plotted as a function of the median fluorescence intensity of tetramer-positive events (Fig. 6, left panel). The strong correlation indicates that individuals with brighter tetramer staining also have higher CD3 competition. T cells that bind greater numbers of tetramer per cell, and are brighter, are interpreted to express a high avidity TCR (14). CD3 competition can be seen to increase both as a function of tetramer brightness and frequency (Fig. 6, right panel). More importantly, CD3 competition can be used to screen out cases with bright tetramer staining at low frequency (Fig. 6, red circles). In this data set, 13 of 61 cases (most of them, the HIV tetramer) fell below the CD3 competition cutoff, despite having frequencies in excess of 1/7,805.
Tetramer staining is widely used today for the detection of antigen-specific T cells. By analyzing the binding of tetramers representing prototype recall and novel peptides and using T cells from HLA-2.1+ and HLA-A2.1− subjects, this report details some of the potential pitfalls and provides guidance for interpretation of results. In addition to describing a new method to determine the specificity of tetramer binding, we have addressed several technical issues that bear on the interpretation of peptide-MHC tetramer results. This is especially important when the population of interest is present at a low frequency and is subject to several sources of artifact that are characteristic of rare event problems (25). Tetramer-positive events were defined in six-parameter space. Because of this, we were able to identify our event of interest as CD3+/CD8+/tetramer positive and eliminate non-T cells, monocytes, debris, and dead cells, which is particularly important when cryopreserved cells are used. CD14 was useful as a “dump” parameter because any event that was positive in FL1, whether a CD14+ monocyte or broadly autofluorescent debris, was eliminated as a potential tetramer-positive event (15). The only alternative to multiparameter flow cytometry is to perform a preseparation step in which sources of interference are eliminated by negative selection (9, 17, 19, 20), with the attendant risk of nonrandom cell loss or alteration. Another technical issue bearing on data analysis addressed here is the use of geometric, rather than arithmetic means to summarize the data and determine cutoffs that define the lower limits of detection. This is analogous to antibody titers, for which the use of geometric mean reciprocal frequency is standard practice (26). Several reports in the current literature have used arithmetic means of percent tetramer-positive cells (6, 9, 27). As a result, mean values are biased by a left shoulder of high frequencies and are therefore overestimated. Similarly, estimates of the sensitivity (lower limit of detection) of the assay would be underestimated. Finally, having applied this methodology, a cohort of HLA-A2.1− individuals was used to define the level of nonspecific tetramer binding, and hence the lower limit of detection of this assay. For FLU and HIV tetramers (as well as others, not shown), the frequency of false CD3+/CD8+/tetramer-positive events was remarkably consistent at approximately 1/20,000 (0.005% positive). Therefore, the limit of detection (the 99th percentile of false-positive events) was estimated at approximately 1/8,000 (0.0125% positive).
We consistently observed that tetramer-positive cells had diminished staining with anti-CD3 mAb, compared to tetramer-negative T cells. It is not likely that the observed diminution in brightness of anti-CD3 was due to internalization, down-regulation, or capping of the CD3/TCR complex. The reason is that staining was performed in the presence of sodium azide and staining with anti-TCR α/β was actually enhanced by tetramer binding. This unexplained, but highly significant, enhancement also rules out the possibility that tetramer-positive cells have inherently lower TCR expression. Further, competition of anti-CD3 binding was dependent on the sequence of staining events: incubation with tetramer prior to anti-CD3 gave the greatest inhibition. In contrast, the frequency of tetramer-positive cells was not significantly affected by the sequence of staining. Finally, for the recall FLU peptide, the magnitude of competition was proportional to the frequency of tetramer-positive cells. This suggests that successive rounds of in vivo exposure to antigen result in selection of T cells that bind peptides more avidly (14) and therefore compete with anti-CD3 antibody binding to a greater extent. This was not the case for the novel HIV antigen. CD45 isoform expression on tetramer-positive cells is consistent with this interpretation as the majority of FLU tetramer-positive cells were CD45RA−/CD45RO+, whereas the majority of HIV tetramer-positive cells were CD45RA+/CD45RO−. Analysis of tetramer brightness alone did not appear to be sufficient to ensure specificity, as a number of cases were identified where tetramer bound brightly at low frequency, but which were CD3-competition negative.
Although the epitope specificity of the commercial antibodies used in this study are not known, competition appeared to be epitope dependent. Binding of one of five anti-CD3 epsilon mAbs was not competed by tetramer. The competition also appeared to be unique to antibodies directed against the CD3 epsilon chain; mAbs directed against CD5 or CD2 were not competed by tetramer binding and, as mentioned, anti-α/β binding was actually enhanced. These data support the idea that CD3 competition requires TCR engagement by tetramer. We suggest that CD3 competition can be used as a marker for tetramer specificity, particularly in cases where the frequency or avidity of tetramer-positive cells is low. Consistent with this interpretation, PBMC from HLA-A2.1− subjects displayed low frequencies of tetramer-positive events (e.g., 1/21,000 for FLU) but failed to compete CD3 binding. This suggests a low level of nonspecific tetramer binding to other sites on the cell surface. We were able to exploit this finding by setting cutoffs for CD3 competition at the upper 99th percentile of competition observed in the HLA-A2.1− cohort (3.2%). By this definition, 93% and 36% of HLA-A2.1+ subjects had detectable frequencies of cells specific for the recall FLU peptide and the novel HIV peptide, respectively. Further, competition of CD3 binding by tetramer appears to be of broader applicability. Our own investigations of tumor-specific immunity in squamous cell carcinoma of the head and neck validated CD3 competition using T-cell lines specific for a tumor-associated peptide (high competition) and an irrelevant peptide (no competition). The assay was then used to evaluate the peptide specificity of T cells freshly isolated from the peripheral blood of patients (unpublished data). Our data indicate that competition of CD3 binding represents a marker for the specificity of tetramer binding to the TCR. Such a marker is needed, as it has been shown that nonspecific tetramer binding may occur by TCR-independent interactions between CD8 and the MHC class I portion of the tetrameric complex (13). In this regard, CD3 competition may prove particularly useful for defining the response to tumor or self- antigens where the frequency of peptide-specific T cells is low and tetramer-positive populations are not always discrete.