|Home | About | Journals | Submit | Contact Us | Français|
A productive CD8+ T-cell response to a viral infection requires rapid division and proliferation of virus-specific CD8+ T cells. Tetramer-based enrichment assays have recently given estimates of the numbers of peptide-major histocompatibility complex-specific CD8+ T cells in naïve mice, but precursor frequencies for entire viruses have been examined only by using in vitro limiting-dilution assays (LDAs). To examine CD8+ T-cell precursor frequencies for whole viruses, we developed an in vivo LDA and found frequencies of naïve CD8+ T-cell precursors of 1 in 1,444 for vaccinia virus (VV) (~13,850 VV-specific CD8+ T cells per mouse) and 1 in 2,958 for lymphocytic choriomeningitis virus (LCMV) (~6,761 LCMV-specific CD8+ T cells per mouse) in C57BL/6J mice. In mice immune to VV, the number of VV-specific precursors, not surprisingly, dramatically increased to 1 in 13 (~1,538,462 VV-specific CD8+ T cells per mouse), consistent with estimates of VV-specific memory T cells. In contrast, precursor numbers for LCMV did not increase in VV-immune mice (1 in 4,562, with ~4,384 LCMV-specific CD8+ T cells per VV-immune mouse). Using H-2Db-restricted LCMV GP33-specific P14-transgenic T cells, we found that, after donor T-cell take was accounted for, approximately every T cell transferred underwent a full proliferative expansion in response to LCMV infection. This high efficiency was also seen with memory populations, suggesting that most antigen-specific T cells will proliferate extensively at a limiting dilution in response to infections. These results show that frequencies of naïve and memory CD8+ T cell precursors for whole viruses can be remarkably high.
The immune response to a viral infection often involves the rapid proliferation of CD8+ effector T cells that recognize virus-infected targets expressing 8- to 11-amino-acid-long peptides on class I major histocompatibility complex (MHC) molecules. This recognition is mediated by membrane-bound T-cell receptors (TCRs) that are generated through largely random DNA recombination events of the many TCRα and -β genes, encoding polypeptide chains that heterodimerize to form the recognition structure of T cells. The recombination of the segments also involves addition or deletion of nucleotides during the joining process, causing even greater diversity, and these processes allow for a very broad range of T-cell specificities, with a calculated theoretical diversity of ~1015 TCRs in the mouse (7). By use of PCR, CDR3 spectratyping, and sequencing techniques, it was estimated that there are approximately 2 × 106 distinct TCR specificities in a mouse spleen (1, 5). This is far below the theoretical level of T-cell diversity, but considering estimates of T-cell degeneracy that propose that a single TCR can recognize up to 106 peptide-MHC (pMHC) complexes (17, 36), it is likely that the functional diversity is much greater than the number of individual TCRs.
It has been of interest to calculate the number of T cells that would either recognize or respond to a pathogen or to a specific pMHC complex. Early estimates of numbers of CD8+ T cells that are specific to a single virus, i.e., precursor frequencies, took advantage of an in vitro limiting-dilution assay (LDA) and calculated CD8+ T-cell virus-specific precursor frequencies to be on the order of 1 in 100,000 in naïve mice and predicted that these cells needed to undergo about 15 divisions to reach the higher precursor frequencies found at day 8 postinfection (29, 30). The efficiency of such assays, however, is relatively poor. Later studies estimated the number of pMHC-specific CD8+ T cells in a naïve mouse by CDR3 sequencing. H-2Kd-restricted T cells specific to HLA residues 170 to 179 (HLA 170-179) were sorted by tetramer from human tumor-immunized mice, and their Vβ CDR3 regions were sequenced. After a plateau suggesting that the majority of the different TCRs had been sequenced was reached, exhaustive sequencing was then used to identify the frequencies of these sequences in naïve mice. These studies found that there were about 600 CD8+ T cells specific for that pMHC complex in naïve mice (4). A second strategy used an in vivo competition assay with H-2Db-restricted lymphocytic choriomeningitis virus (LCMV) GP33-specific P14-transgenic T cells to estimate the number of GP33-specific CD8 T cells in naïve mice and calculated the number to be between 100 to 200 cells per mouse (2).
Others estimated numbers of pMHC-specific T cells by sequencing the CDR3β regions of antigen-specific T cells that had expanded during an acute infection. By calculating a measure of CDR3 diversity and then assuming a logarithmic distribution of diversity, they extrapolated the number of T-cell clones that responded to an acute infection. With this technique, 300 to 500 H-2Db-restricted mouse hepatitis virus (MHV)-encoded S510 clonotypes were calculated to be in the central nervous systems of acutely infected mice, with ~100 to 900 clonotypes calculated to be in chronically infected mice (24). Later studies used a gamma interferon (IFN-γ) capture assay instead of tetramer sorting and estimated 1,100 to 1,500 H-2Db-restricted S510-specific clonotypes and 600 to 900 clonotypes of the subdominant H-2Kb-restricted MHV S598 peptide-specific T cells in the spleens of acutely infected mice (25). Those studies also estimated that there were 1,000 to 1,200 different H-2Db-restricted GP33-specific clonotypes that could respond to an LCMV infection.
More-recent studies have taken advantage of magnetic tetramer binding enrichment and double tetramer staining of cells from the spleen and lymph nodes of naïve mice to determine pMHC precursor frequencies, with the assumption that most CD8+ T cells in a naïve mouse reside in lymphoid organs and will react with tetramers. This technique was first described by Moon et al. for CD4+ T cells, and it detected ~190 I-Ab 2W1S 52-68-specific T cells, ~20 I-Ab Salmonella enterica serovar Typhimurium FLiC 427-441-specific T cells, and ~16 I-Ab chicken ovalbumin (OVA) 323-339-specific T cells per mouse (19). This same technique was then used to determine numbers of pMHC-specific CD8+ T cells for epitopes derived from a variety of viruses and found 15 to 1,070 pMHC-specific CD8+ T cells per mouse, depending on the specificity of the pMHC tetramer (10, 15, 23). Determinations of CD8+ T-cell precursor frequencies in humans are currently not experimentally attainable, but exhaustive sequencing of an HLA-A2.1-restricted influenza A virus (IAV) M1 58-66-specific T-cell response has suggested that there are at least 141 different clonotypes that can grow out in response to an in vitro stimulation with peptide, providing a minimum number of T cells that can respond to this pMHC complex in humans (22).
Most of the assays estimate the number of T cells specific to single peptides in individual mice. These assays, therefore, do not determine the numbers of CD8+ T cells that can proliferate in response to an entire virus, especially if the virus is known to have many epitopes or if epitopes for the virus have not been described. By examining the average number of pMHC-specific CD8+ T cells in a naïve mouse and comparing this to the number of pMHC-specific CD8+ T cells that are in a mouse at the peak of the T-cell response, it can be calculated that CD8+ T cells divide approximately 12 to 14 times after virus infection (23). Considering that the progeny of one precursor after only 12 divisions can result in just over 4,000 cells, and since recent experiments using H-2Kb-restricted chicken OVA 257-264-specific OT-1-transgenic T cells have confirmed that the progeny from a single cell can be detected in a mouse after infection (31), an in vivo LDA was set up to take advantage of the extensive division and proliferation of virus-specific CD8+ T cells in order to determine virus-specific CD8+ T-cell precursor frequencies.
Here, we show that by transferring limiting amounts of carboxyfluorescein succinimidyl ester (CFSE)-labeled Thy1.1+ Ly5.2+ heterogeneous CD8+ T cells into Thy1.2+ Ly5.1+ hosts, we are able to calculate CD8+ T-cell precursor frequencies for whole viruses. Our calculations are based on finding the number of donor CD8+ T cells that results in low-level-CFSE (CFSElo) (i.e., proliferated) donor CD8 T cells in 50% of the hosts. Using probit or Reed and Muench 50% endpoint calculations (3, 26), we are able to calculate CD8+ T-cell precursor frequencies. We show here that frequencies of naïve CD8+ T-cell precursors for whole viruses are quite high and that our in vivo LDA calculates whole-virus precursor frequencies in line with determinations using other methods with naïve and immune mice.
B6.SJL (Ly5.1+ Thy1.2+ host) mice were used between 6 and 20 weeks of age and were either obtained from Taconic Farms (Germantown, NY) or bred in our own mouse-breeding colony. B6.Cg-IgHa Thy-1a GPi-1a/J (Ly5.2+ Thy1.1+ donor) mice were used at 6 to 32 weeks of age and bred in our own mouse-breeding colony. Transgenic TCR-LCMV-P14 mice were used at 6 to 20 weeks of age and bred in our own mouse-breeding colony. All experiments were done in compliance with the Institutional Animal Care and Use Committee of the University of Massachusetts Medical School (Worcester, MA).
LCMV strain Armstrong was propagated in BHK21 baby hamster kidney cells (34, 38). Vaccinia virus (VV) strain Western Reserve was propagated on L929 cells (38). Mice were inoculated intraperitoneally with 5 × 104 PFU of LCMV or 1 × 106 PFU VV in 0.2 ml for acute viral infections, and some of these mice were tested for memory T-cell responses 3 to 6 months later.
CFSE labeling of splenocytes was performed as previously described (9, 16). Briefly, a single-cell suspension was prepared from the spleen, red blood cells were lysed in an 0.84% NH4Cl solution, and splenocytes were washed in cold Hank's balanced salt solution (HBSS) (Gibco ; Invitrogen, Carlsbad, CA) and resuspended in HBSS for counting. Spleen leukocytes were then resuspended in a 2 μM CFSE solution in HBSS at 2 × 107 per ml and labeled for 15 min in a 37°C water bath, with mixing every 5 min. After CFSE labeling, cells were again washed twice with cold HBSS and counted immediately before transfer. An aliquot of splenocytes was used for a surface stain, and the rest of the splenocytes were diluted in HBSS for adoptive transfer.
Single-cell suspensions of the spleen, lymph nodes, bone marrow, blood, peritoneal cavity, and lungs (minced finely with a razor blade and filtered) were prepared; red blood cells were lysed in an 0.84% NH4Cl solution; and the leukocytes were then washed in RPMI 1640 medium (11875-093; Sigma-Aldrich, St. Louis, MO). Cells were then counted by a hemacytometer and resuspended in fluorescence-activated cell sorting (FACS) buffer for staining. Fc receptors were blocked with antibody to CD16/CD32 (Fcγ III/II receptor ; BD Biosciences, San Diego, CA), and cells were then stained in 96-well plates. For the in vivo LDA, the single-cell suspension from each whole spleen was divided into 8 to 16 wells of a 96-well plate for staining and later recombined for analysis on an LSRII flow cytometer. After the surface stain with the indicated antibodies, cells were either fixed using Cytofix (554655; BD) and resuspended in FACS buffer for analysis or, for intracellular assays, permeabilized using Cytofix/Cytoperm (554722; BD) and stained intracellularly with the indicated antibodies per the manufacturer's instructions.
CD3 phycoerythrin (PE)-Cy7 (552774; BD), CD8α Pacific Blue (558106; BD), Thy1.1 PE (554898; BD), Ly5.2 peridinin chlorophyll protein (PerCP)-Cy5.5 (552950; BD), and Vα2 allophycocyanin (APC) (17-5812-80; eBioscience) were used for in vivo LDAs. For the intracellular cytokine assay, monoclonal antibody (MAb) to CD8α Alexa Fluor 700 (557956; BD), CD44 PE-Cy7 (25-0441-82; eBioscience, San Diego, CA), Thy1.1 PE (554898; BD), and Ly5.2 PerCP-Cy5.5 (552950; BD) were used for the surface stain, and MAb to IFN-γ APC (554413; BD) was used for the intracellular stain. Peptides for stimulation were purchased from 21st Century Biochemicals (Marlboro, MA). For comparison of uninfected donor and host T-cell phenotypes, CD3 PE Cy7 (552774; BD), CD8α Alexa Fluor 700 (557956; BD), Thy1.1 PE (554898; BD), CD127 APC (17-1271-82; eBioscience), CD62L Pacific Blue (57-0621-82; eBioscience), and CD44 PerCP-Cy5.5 (45-0441-82; eBioscience) were used.
Peptide stimulations were performed as previously described (33). Briefly, single-cell suspensions of lymphocytes were cultured for 5 hours in the presence of 3 μM of the indicated peptides (21st Century Biochemicals) or purified MAb to CD3 (1 μg/ml) (553058; BD) for a polyclonal stimulation, with human recombinant interleukin-2 (10 U/ml) and GolgiPlug (555029; BD).
The take of Ly5.2+ Thy1.1+ donor CD3+ CD8+ cells in Ly5.1+ Thy1.2+ host mice was determined by plotting the log10 of the number of donor CD3+ CD8+ cells transferred by the log10 of the number of CD3+ CD8+ donor cells recovered in the spleens of uninfected host mice. Mice that received fewer than 1.25 × 105 splenocytes were not included in the analysis, because at this number transferred, donor CD3+ CD8+ cells were not reproducibly detectable in uninfected hosts. The resulting formula was then used to calculate a percent take in the spleen. Then, using the assumption that 67% of all CD3+ CD8+ cells in a naïve uninfected mouse reside in the spleen (6, 8), we were able to calculate a total donor take.
Splenocytes were labeled with CFSE as described above and diluted in HBSS to appropriate concentrations. Pilot experiments gave an indication as to the limiting number of T cells that would need to be in host mice to respond to each viral infection. To determine precursor frequencies, twofold dilutions of splenocytes were made in HBSS. Each dilution was transferred into four to six (usually five) host mice. In each experiment, one mouse at each dilution was left uninfected to serve as a negative control. An infected mouse was scored as a responder if donor Thy1.1+ Ly5.2+ CFSElo cells were detected after FACS analysis to be above a determined threshold as described below. If there were Thy1.1+ Ly5.2+ CFSElo cells in any dilution of any of the uninfected animals, this number of cells was multiplied by three (with a correction for number of cells collected), and this would serve as a responder cutoff. In instances where there was no background detected by FACS in the uninfected mice of the individual experiment, a cutoff of 10 CFSElo cells was used. The background value of 10 CFSElo cells was used because it was three times the average number of CFSElo cells detected in all uninfected mice in all experiments, and it was also the average number of CFSElo cells in uninfected mice that had detectable CFSElo cells plus 1 standard deviation. After determination of responders versus nonresponders, probit and Reed and Muench 50% endpoint analyses were performed, and the resulting number was multiplied by two to determine the limiting number of transferred cells required to result in a responder ~100% of the time, i.e., the precursor frequency (3, 26). These two analyses resulted in comparable but not identical precursor numbers. Almost all individual experiments also included two control host mice that received adoptive transfers of a large number of donor splenocytes, with one mouse infected and one uninfected, serving as positive and negative controls.
An in vivo LDA was designed to estimate the CD8+ T-cell precursor frequency for entire viruses. This was achieved by adoptive transfer of donor splenocytes into host mice that differed by using two congenic markers (Ly5 and Thy1) to decrease the fluorescent background when donor CD8+ T-cell populations were stained for and by increasing the detection limit of resultant T-cell progeny by counting only CD8+ events by FACS. By ignoring non-CD8+ events, we increase the total number of CD8+ events that we were able to collect, and we could collect just over 4 × 106 CD8+ events, about one-fifth of the total number of CD8+ T cells in an uninfected animal (assuming 2 × 107 CD8 T cells per mouse) and, because of T-cell proliferation, approximately 5% of all CD3+ CD8+ events in an LCMV- or VV-infected mouse. This allows reliable detection of donor CD8+ T-cell progeny at limiting dilutions.
To set up the in vivo LDA, we verified that the adoptive transfer of donor cells into host mice did not alter the phenotype of donor cells and that adoptive transfer of decreasing numbers of T cells resulted in a linear decrease of donor T cells in host mice. B6.Cg-IgHa Thy-1a GPi-1a/J (Ly5.2+ Thy1.1+ donor) splenocytes were labeled with CFSE and diluted so that each mouse would receive ~5 × 107 splenocytes. Twofold dilutions of this stock were made, and these samples were transferred intravenously into groups of four B6.SJL (Ly5.1+ Thy1.2+ host) mice. After 3 days, mice were sacrificed and immunophenotyping of the splenocytes was performed. Donor CD8+ T cells had phenotypes that remained largely naïve, with high levels of CD127 and CD62L and mostly low levels of CD44, and were similar to the phenotype of host CD3+ CD8+ T cells (Fig. (Fig.1A).1A). As decreasing numbers of T cells were transferred into host mice, the number of donor CD8+ T cells detected in the spleen linearly decreased, with an R2 value of 0.994 (Fig. (Fig.1B1B).
To test whether adoptively transferred T cells would traffic normally throughout the body, Ly5.2+ Thy1.1+ donor splenocytes were labeled with CFSE, and 1.35 × 107 of these cells was transferred into Ly5.1+ Thy1.2+ hosts. After 5 days, mice were sacrificed, and FACS analysis of cells from lymphoid organs was performed. We found that Ly5.2+ Thy1.1+ donor CD8+ T cells trafficked at similar frequencies to the mediastinal lymph nodes, the axillary lymph nodes, the peribronchial lymph nodes, the spleen, and, to some extent, the bone marrow (Fig. (Fig.2A).2A). When total numbers of recovered Ly5.2+ Thy1.1+ donor CD8+ T cells in these tissues were counted, 65% of the recovered donor CD8+ T cells were in the spleen. In a separate, similar experiment, peripheral sites were examined. There was a reproducibly detectable number of donor T cells in peripheral sites such as the peritoneal cavity and lungs, although the take in such peripheral sites was lower than that in lymphoid sites such as the spleen or mediastinal lymph nodes (Fig. (Fig.2B2B).
To ensure that transferred T cells had immunodominance hierarchies comparable to the ones observed in normal C57BL/6J mice, we transferred large numbers (~5 × 107) of CFSE-labeled Ly5.2+ Thy1.1+ donor splenocytes into Ly5.1+ Thy1.2+ host mice, waited 3 days, and infected the mice with LCMV. On day 7 of LCMV infection, intracellular IFN-γ assays were performed by stimulating spleen cells with LCMV peptides or by polyclonal stimulation with MAb to CD3. The percentages of host and donor CD8+ T cells that produced IFN-γ when stimulated with MAb to CD3 or with LCMV-specific peptides GP33, NP396, GP276, GP118, and NP205 were similar in donor and host CD8+ T cells (Fig. (Fig.3).3). Comparable experiments were performed using VV, and the hierarchies of CD8+ T cells that produced IFN-γ after polyclonal or VV-specific peptide stimulation were also similar in donor and host CD8+ T cells (CD3 > B8R > A47L) (data not shown).
To determine virus-specific CD8+ T-cell precursor frequencies, graded amounts of splenocytes were transferred into host mice at limiting-dilution numbers that result in donor proliferated CD8+ T cells in ~50% of hosts. Ly5.2+ Thy1.1+ donor splenocytes from uninfected mice were labeled with CFSE and transferred at decreasing numbers (5 × 106, 2.5 × 105, 1.25 × 105, and 0.625 × 105) into Ly5.1+ Thy1.2+ hosts. Pilot experiments had demonstrated that in all host mice at the 5 × 106 dose, some donor CD8+ T cells proliferated, as shown by a CFSElo cell peak in response to a VV infection, so this dilution was used in subsequent experiments as a positive control. The 2.5 × 105-, 1.25 × 105-, and 0.625 × 105-splenocyte dilutions resulted in responders and nonresponders, so these dilutions were used for in vivo LDA calculations. A small aliquot of the transferred splenocytes was stained, and FACS analysis was performed to determine the exact number of CD8+ T cells transferred into host mice. Previous experiments (data not shown) indicated that VV-specific CD8+ T-cell responses peaked at day 6, so on day 6, mice were sacrificed and their splenocytes analyzed by FACS under two conditions. For each spleen, a small aliquot was run, analyzing all events to determine the CD3+ CD8+ percentage, and then, to allow detection of the small number of donor CD3+ CD8+ progeny at limiting dilutions, the rest of the spleen was analyzed, with the threshold set to collect only CD8+ events. Figure Figure4A4A is an example of a transfer at limiting dilution in VV-infected or uninfected mice. Responders versus nonresponders were scored as described in Materials and Methods. Figure Figure4B4B is an example of an in vivo LDA for VV, where the numbers of responders per concentration are three of four at the high concentration, two of four at the intermediate concentration, and one of four at the low concentration. Figure Figure4C4C is an example of an in vivo LDA using Ly5.1+ H-2Db-restricted LCMV GP33-specific P14-transgenic T cells (P14-transgenic T cells) transferred into C57BL/6J (Ly5.2+) mice subsequently infected with LCMV (Fig. (Fig.4C).4C). In this experiment, the numbers of responders per concentration were three of four at the high concentration, one of four at the intermediate concentration, and zero of four at the lowest concentration.
One caveat for determination of precursor frequencies by use of an adoptive transfer method involves the donor take. Not all adoptively transferred donor CD8+ T cells survive in the host, and the percentage that does survive in a host mouse has been referred to as the donor take. Figure Figure55 is a graph of data from 91 uninfected mice and plots the number of Ly5.2+ Thy1.1+ donor CD8+ T cells transferred into Ly5.1+ Thy1.2+ hosts against the number of Ly5.2+ Thy1.1+ donor CD8+ T cells found in the spleen. These data give us a formula that allows us to determine a splenic donor take. This take, along with the assumption that approximately 67% of CD8+ T cells reside in the spleen, a number calculated by others (6, 8), allows us to estimate a full mouse CD8+ T-cell take of approximately 3.8% at low cell numbers. This may be a slight underestimate in comparison to our own results suggesting that less than 65% of donor CD3+ CD8+ cells reside in the spleen (Fig. (Fig.22).
Using multiple in vivo LDAs, our take value, and probit analysis for 50% endpoint times 2 (3), we determined T-cell precursor frequencies in naïve and immune mice as shown in Table Table1.1. There were about 1 in 2,958 ± 392 CD8+ T cells in naïve mice that proliferated in response to LCMV, while there were almost twice as many CD8+ T cells, 1 in 1,444 ± 171, that proliferated in response to VV (P < 0.0001 in comparison to LCMV precursors). The number of CD8+ T cells in VV-immune mice that proliferated in response to VV was, as expected, greatly increased, with 1 in 13 ± 2 CD8+ T cells able to proliferate in response to this homologous infection (P < 0.0001 in comparison to the naïve immune state). As expected, the LCMV-specific CD8+ T-cell precursor frequency was not elevated in VV-immune mice; in fact, it was slightly, although significantly, decreased, with about 1 in 4,425 ± 1,705 CD8+ T cells proliferating (P < 0.05 in comparison to the naïve immune state).
We calculated the number of P14-transgenic T cells that responded to an LCMV infection by our in vivo LDA and determined a frequency by probit analysis of 1 in 0.93 ± 0.04. This suggests a virtually 100% efficiency in the outgrowth of the transgenic T cells and reinforces the calculations that we have made concerning T-cell take after transfer.
We also employed the commonly used Reed and Muench 50% endpoint analysis to determine precursor frequencies (26) and found that these calculations resulted in precursor frequencies comparable but not identical to those obtained with the probit method. By Reed and Muench analysis, naïve mice had 1 in 3,121 ± 291 CD8+ T cells specific to LCMV and 1 in 1,615 ± 409 CD8+ T cells specific to VV, while in VV-immune mice, 1 in 3,956 ± 787 were specific to LCMV, and 1 in 13 ± 1 were specific to VV in these immune mice. Using the Reed and Muench method, we calculated that 1 in 1.22 ± 0.11 P14-transgenic T cells responded to an LCMV infection, again a figure close to 100%.
The broad possible pMHC reactivity generated by random gene rearrangements of α and β TCR chains on T cells would seem to ensure reactivity against a diverse array of pathogens, but this broad diversity then raises the question of how many T cells in a host would respond to a specific pMHC complex or against an entire pathogen. Many determinations of CD8+ T-cell precursor frequencies have now been used to calculate the number of pMHC-specific CD8+ T cells within a mouse (Table (Table2),2), but the determination of the total number of CD8+ T cells that could respond to a viral infection would be possible with these methods only if all of the epitopes of the virus were known. However, the numbers of epitopes found to stimulate T-cell responses are now becoming quite large, as means for detecting them have become more sensitive. A virus like VV is now reported to encode close to 50 H-2Kb- and H-2Db-restricted epitopes (20), a number that would make it very difficult, if not prohibitive, if the precursor frequency for the entire virus was to be determined with the usual methods. Herein, we have described an in vivo LDA that allows an unbiased determination of precursor frequencies for entire viruses without any knowledge of the specificity or number of epitopes.
It is interesting to note the differences calculated for T-cell precursor frequencies depending on the method used (Table (Table2).2). Extrapolation of Vβ clonotype number per spleen by calculation of a measure of diversity by examination of the CDR3 sequences of pMHC-specific populations as done previously (12, 24, 25) gives precursor frequencies on the high ends of most estimates, with the highest number of H-2Db LCMV GP33-specific CD8+ T cells calculated among all methods. This is interesting in that this method is assumed to be an underestimate, because it does not include TCRα diversity, and it also does not account well for redundancy in T-cell populations, i.e., there may be more than one cell of a single clone in a naïve mouse. However, there is some uncertainty in these numbers because they rely on extrapolations of the numbers and diversities of sequenced clones to estimate T-cell precursor frequencies. Compared to our results, those obtained with this extrapolation method would suggest our calculations to be slight underestimates. If there are 1,100 to 1,200 GP33-specific CD8+ T cells per naïve mouse spleen, and the GP33-specific CD8+ T-cell response is approximately 10% of the total LCMV-specific response, we would expect to find about twice as many LCMV-specific precursors. Instead of the 6,760 per mouse as we calculated, we would expect 11,000 to 12,000 LCMV-specific CD8+ T cells.
If precursor frequencies are instead calculated by transgenic-T-cell competition (2), where transferred monoclonal transgenic T cells compete against heterogeneous endogenous T-cell populations to determine precursor frequencies, our results look to be overestimates. Assuming a 10% take and 2 × 107 CD8+ T cells per mouse, this method estimates ~100 H-2Db LCMV GP33-specific CD8+ T cells per mouse (2). This result would put our LCMV-specific CD8+ T-cell precursor determination on the high end.
The tetramer-based enrichment assay, which makes use of pMHC tetramers, magnetic-bead enrichment and double-tetramer FACS staining of spleens and lymph nodes to identify pMHC-specific CD8+ T cells (assuming that most naïve T cells reside in lymph organs), seems to yield numbers that are in line with results determined by our in vivo LDA. Depending on the individual determination, there are ~287 (23) or ~449 (15) H-2Db LCMV GP33-specific CD8+ T cells in all lymph organs of a naïve mouse, and this result would be on the low end yet still compatible with what our results might predict for a frequency of GP33-specific naïve CD8+ T cell precursors. For VV, the in vivo LDA calculates about 13,850 responsive CD8+ T cells per C57BL/6J mouse. The tetramer-based enrichment assay estimated 1,070 CD8+ T cells specific for the H-2Kb-restricted VV B8R epitope in the spleen, lymph nodes, and ovaries (10), and those results would seem consistent with the results that we have described, considering that the B8R peptide response may represent about 10% of the VV-induced response.
About twice as many T cells were responsive to VV than to LCMV (P < 0.0001). This might in part reflect the observations that the T-cell response to VV peaks earlier than that of LCMV. Having more CD8+ T cells that are specific to VV may increase the likelihood that VV-specific CD8+ T cells interact with stimulating antigen-presenting cells earlier, allowing peak T-cell proliferation to occur earlier. VV also encodes more proteins than LCMV, with almost twice as many VV epitopes described to occur in the C57BL/6J mouse, consistent with the result showing almost twice as many VV-specific precursor T cells than LCMV-specific CD8+ T cells (14, 20).
Our results also estimate that 8% of CD8+ T cells in VV-immune mice are VV responsive, and these data are supported by results obtained from VV-immune mice by using peptide stimulations and intracellular cytokine stains that estimate that anywhere from 2 to 11% of CD8+ T cells are specific to the VV-encoded immunodominant B8R epitope in D21 or D40 VV-immune mice (28, 32), and our own results from intracellular cytokine assays estimate that 0.5 to 2% of CD8+ T cells in VV-immune mice are specific to the VV B8R epitope at 3 to 8 months postinfection (data not shown). This increase in CD8+ T-cell precursor frequency for VV in VV-immune animals by more than 2 orders of magnitude (P < 0.0001) demonstrates the expected considerable increase of VV-specific memory CD8+ T cells after VV infection. As expected, there was no increase in the number of CD8+ T cells that respond to LCMV in VV-immune mice, and this helps to validate the specificity of our assay. The small but significant decrease in LCMV CD8+ T-cell precursor frequency in VV-immune mice is interesting and may suggest that memory cells may displace some naïve cells in the immune response. We have not systematically addressed changes in VV-specific precursors in LCMV-immune mice because there is a high degree of heterologous immunity in this virus sequence, and the immunity, due to private specificities in the immune repertoire, has such high variability that our in vivo LDA would likely suffer from reproducibility issues (13).
It is possible to make an approximation of the number of divisions a CD8+ T cell undergoes after stimulation by examining the burst size or recovered cell number at the limiting dilution. By determining the frequency of CFSElo donor cells among all CD8+ events collected, multiplying that frequency by the total number of CD8+ T cells found in the spleen, and then multiplying that number in accordance with the assumption that 67% of all CD8+ T cells are present in the spleen during infection, we are able to calculate the approximate number of divisions a CD8+ T cell undergoes after virus infection. The numbers of divisions that a VV-specific precursor undergoes by day 6 (~11 divisions) and that an LCMV-specific precursor undergoes by day 7 (~12 or 13 divisions) fall within predicted ranges. However, we approximate that a P14-transgenic T cell undergoes ~14 divisions by day 7 of an LCMV infection, and this is significantly different (P = 0.023) from the number (~12 or 13 divisions) that a naïve CD8+ T cell from a heterogeneous population of T cells undergoes. This may reflect differences in avidity between the transgenic T-cell population and the expected large range of avidities of T cells in a heterogeneous population as a whole or may instead be related to the examination of a monoclonal T-cell population that responds to a highly expressed immunodominant epitope versus a heterogeneous population of CD8+ T cells that contains T cells responding to immunodominant and subdominant epitopes.
The immunological environment produced by a specific virus infection can have a profound impact on the burst sizes of epitope-specific T cells, as has been demonstrated by experiments examining the T-cell response to recombinant viruses engineered to express the same T-cell epitope (21). One explanation for this would be the expression of insufficient antigen to engage all of the T-cell precursors, as shown previously (11, 18). However, we used doses of virus that maximized the burst of the T-cell response, as in Ref (11), and we feel that virtually all the precursors should have been engaged. Finally, T cells of differing affinities may undergo different numbers of divisions and expand to different peak sizes, as recently demonstrated (39). Since the in vivo LDA requires extensive proliferation, it is possible that we have missed lower-affinity clones in our assay that would divide fewer than seven or eight times, and this would make our calculated precursor frequencies underestimates. However, analyses of T-cell responses to several epitopes suggest at least 12 to 14 divisions per T cell (23), which would be detected by our assays. Further, our in vivo LDA seems to be able to detect the bulk of memory T-cell precursors detected by intracellular IFN-γ assays.
Our calculated take of CD8+ T cells is not the normally quoted 10% figure. If we instead use a 10% value for take, the precursor frequencies would be decreased with a naïve mouse having about 1 in 7,805 ± 1,034 CD8+ T cells specific for LCMV and 1 in 3,809 ± 452 CD8+ T cells specific for VV and a VV-immune mouse having about 1 in 34 ± 6 and 1 in 12,036 ± 4,812 CD8 T+ cells specific for VV and LCMV, respectively. Our own rough estimate of total cell numbers (Fig. (Fig.2)2) would suggest that slightly less than 67% of CD8+ T cells reside in the spleen, as had been suggested (8), but given that our attempt to count T cells throughout the body was not exhaustive and that the take would probably only change by at most a 20 to 30% value, we remain confident that our calculations are within a reasonable range of total virus-specific CD8+ T-cell precursor frequencies. Further, our experiments using P14-transgenic T cells also strongly support our estimation of take since our calculations of numbers of precursors equals the number of transgenic T cells obtained using that take value. The efficiencies of T-cell take in our experiments may seem to be in conflict with studies by others using OT-1-transgenic T cells, where 25% of the mice injected with a single OT-1-transgenic T cell had detectable responding donor T cells (31). However, these single cells were injected intraperitoneally, followed by an immediate intraperitoneal infection. We, instead, chose to allow for a total-body distribution of T cells by way of an intravenous transfer and challenged mice with virus 2 to 5 days later. We therefore remain confident of our take value under the conditions of our system. Further, this in vivo LDA calculated at a 3.8% take gives frequencies with high concordance with the anticipated number of VV-specific memory cells, which can be measured directly by intracellular-cytokine assays. Considering the large amount of data we have in generating the 3.8% figure (Fig. (Fig.5),5), we believe that our estimate of take is reasonably accurate in these experiments.
The in vivo LDA could be used to examine virus-specific T-cell precursor frequencies in mice of different ages or physiological states or in mice with histories of different infections. For example, a decline in IAV-specific repertoire diversity leading to epitope-specific holes in the repertoire in aged mice was recently reported (37). Future experiments may be able to use the in vivo LDA described herein to determine whether whole-virus T-cell precursor frequencies in naïve or immune aged mice are also changed.
Whereas tetramer-based enrichment assays measure the number of cells that are reactive to a particular pMHC complex, the in vivo LDA requires CD8+ T-cell division and proliferation, measuring instead the number of CD8+ T cells that do proliferate in response to a viral infection. It might seem likely that not all virus-specific T cells would react with tetramer and that not all tetramer-specific T cells would be capable of proliferating in response to their cognate antigen. Remarkably, though, while the in vivo LDA described here tells us something different than the tetramer-based enrichment assays described recently, it gives reasonable concordance with those techniques.
This work was supported by U.S. National Institutes of Health research grants U19-AI-057330, RO1-AR-35506, and R37-AI-17672 and training grant T32 AI07349.
Published ahead of print on 7 October 2009.