INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Virus-specific cytotoxic CD8⁺ T cells (CTL) play a central role in the immune response to some virus infections by eliminating virus-infected cells (40). Control of the primary infection is followed by establishment of a "memory" population that is able to respond more rapidly and vigorously to a secondary infection with the same virus (12, 40). Recently, the development of major histocompatibility complex (MHC) class I tetramers has made it possible to directly identify antigen-specific T cells ex vivo (3). These studies have revealed that memory CD8⁺ T cells are maintained at unexpectedly high frequencies in the spleen. For example, intraperitoneal infection of mice with lymphocytic choriomeningitis virus (LCMV) induces antigen-specific memory CD8⁺ T cells at frequencies as high as 8% of splenic CD8⁺ T cells (6, 29). This is much higher than the frequencies detected by classical limiting-dilution assays (LDA). Relatively high frequencies of CD8⁺ memory T cells have also been detected following influenza virus infection (0.5% CD8⁺ T cells) (16), although these frequencies are significantly lower than those induced by LCMV infection.

Memory cells have classically been considered to have a resting phenotype; however, evidence is now emerging that some memory cells can respond rapidly to antigen and may maintain an effector function (4, 6, 19, 24, 29, 30, 32). Recent studies have revealed that there is substantial heterogeneity among populations of memory cells with respect to cell turnover (33) and phenotype as measured by activation markers such as CD62L and CD45RA/B/C (13, 33). In addition, there is evidence of functional heterogeneity in persistent LCMV infection, where there is a population of antigen-specific CD8⁺ T cells which express activation markers and proliferate in vivo but cannot exert an effector function (41). Other studies have suggested that a CD62L^hi population of memory CD8⁺ T cells which lose their immediate effector function can be induced in primary culture (30). However, these cells were generated following primary activation of T-cell receptor transgenic T cells in vitro and it is not clear how this relates to the situation with memory cells generated in response to a natural pathogen.

Sendai virus is a type 1 parainfluenza which is a natural respiratory pathogen of mice. Intranasal infection of mice with Sendai virus elicits a potent CD8⁺ T-cell response that is almost exclusively directed at a single K^b-restricted nucleoprotein (NP) epitope defined by a nonapeptide, NP_324-332 (9, 10, 20, 23). There is also a response to a subdominant epitope generated by the same NP_324-332 peptide presented in the context of D^b. Although CTL specific for the NP_324-332/D^b epitope do not take part in the primary effector response, stable long-term memory for both NP_324-332/K^b and NP_324-332/D^b epitopes can be detected by LDA (8, 15). In the current study, we took advantage of MHC tetramers to further investigate the CD8⁺ memory T cells elicited by Sendai virus infection of mice. The data show that Sendai virus induces a high frequency of memory CD8⁺ T cells, which is much higher than that detected by LDA. Furthermore, there is significant heterogeneity among the CD8⁺ memory T cells with respect to proliferative capacity, and this is associated with CD62L expression.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Mice and virus. Female C57BL/6 mice were purchased from Jackson Laboratories (Bar Harbor, Maine) and housed under specific-pathogen-free conditions. Mice were infected at 6 to 12 weeks of age. The Enders strain of Sendai virus was grown, stored, and titrated as described previously (20). Mice were anesthetized by intraperitoneal injection with Avertin (2,2,2-tribromoethanol) and infected intranasally with 500 50% egg-infectious doses of Sendai virus. Influenza virus A/HK-x31 (H3N2) was grown, stored, and titrated as previously described (11). Mice were infected intranasally under anesthesia with 240 hemagglutination units of virus. Mice were considered to be "memory mice" when they had been infected with Sendai virus a minimum of 30 days previously.

Peptides. Sendai virus NP_324-332 and influenza virus NP_366-374 (38) peptides were synthesized at the St. Jude Children's Research Hospital Center for Biotechnology by using a Perkin-Elmer Applied Biosystems (Berkeley, Calif.) 433A peptide synthesizer. Peptide purity was evaluated by reverse-phase high-pressure liquid chromatography analysis.

Carboxyfluorescein (diacetate) succinimidyl ester (CFSE) labeling and culture conditions. Spleen cells were labeled with CFSE by incubation with 0.5 to 0.7 µM CSFE diluted in Hanks balanced salt solution for 10 min in the dark. Cells were subsequently washed with Hanks balanced salt solution or complete tumor medium (22) before use. Unless otherwise stated, cultures were restimulated in vitro with NP_324-332 peptide at a concentration of 0.5 µg/ml and human recombinant interleukin-2 (IL-2) at 10 U/ml (R&D Systems, Minneapolis, Minn.) at a cell density of 10⁶/ml in 24-well plates.

BrdU labeling. Spleen cells were restimulated in vitro with NP_324-332 peptide as described above and cultured in the presence of 10 µM BrdU and 1 µM fluorodeoxyuridine FdUrd (14). After 48 h, cells were harvested and stained with NP_324-332/K^b tetramer plus anti-CD8 and anti-BrdU antibodies in accordance with established protocols (37).

MHC tetrameric reagents and analysis. The construction of folded MHC class I-peptide complexes and their tetramerization have been described previously (29). Four tetramers were used. They were K^b folded with Sendai virus NP_324-332 (FAPGNYPAL), D^b folded with the same peptide, and as controls, D^b folded with influenza virus NP_366-374 (ASNENMETM) and K^b tetramers folded with the peptide TSINFVKI (p79) derived from murine gammaherpesvirus 68 (MHV-68) (35). Tetramers were stored as aliquots either at 70°C or at 4°C. The titer of the Sendai virus NP_324-332/D^b tetramer was determined by using an NP_324-332/D^b-specific CD8⁺ T-cell line, and that of the Sendai virus NP_324-332/K^b tetramer was determined by using bronchoalveolar lavage (BAL) fluid from Sendai virus-infected C57BL/6 mice. The influenza virus NP/D^b peptide tetramer and murine MHV-68/K^b acted as negative controls for the tetramers folded with Sendai virus epitopes; no cross-reactivity between the tetramers was detected. Staining with tetrameric reagent took place for 1 h at room temperature, followed by staining with anti-CD8 tricolor (Caltag, Burlingame, Calif.) and fluorescein isothiocyanate-conjugated anti-CD44 or anti-CD62L (Pharmingen, San Diego, Calif.) on ice for 20 min. For four-color staining of in vitro-restimulated cultures, CFSE-labeled cells were stained with allophycocyanin-conjugated anti-CD8 (Caltag), phycoerythrin-conjugated NP_324-332/K^d tetramer, and biotinylated anti-CD44, anti-CD62L, anti-Ly6C, anti-CD25, or anti-CD69 (Pharmingen), followed by phycoerythrin-Cy7-conjugated streptavidin (Caltag). Three-color-stained samples were run on a Becton-Dickinson FACScan flow cytometer, and four-color-stained samples were run on a FACSCalibur. Data were analyzed by using CELLQuest software (Becton Dickinson Immunocytometry Systems, San Jose, Calif.). In some experiments, B cells were depleted before staining by panning on anti-immunoglobulin-coated flasks.

Cell sorting. Cell preparations were stained with the NP_324-332/K^b tetramer and/or antibodies as described above and then sorted into the appropriate cell populations by using either a FACSstar Plus or a MoFlo cell sorter (Becton Dickinson and Cytomation, respectively). Sorted cell populations were generally >90% pure.

CTL assays. L-K^b target cells have been described previously (7, 31). Cells were loaded with peptide and Na⁵¹CrO₄ as described previously (8). Briefly, monolayers of L-K^b cells were incubated with 150 µCi of Na⁵¹CrO₄ with or without the NP_324-332 peptide at approximately 300 µg/ml in a minimal volume for 1 h. One milliliter of complete tumor medium was added, and the targets were incubated overnight. Cells were washed and counted, and standard protocols were used for ⁵¹Cr release assays (2). The percentage of specific release was calculated by using the formula % specific release = (experimental  spontaneous)/(maximum  spontaneous). Spontaneous release was typically <10% of maximum release.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

High frequencies of memory CD8⁺ T cells specific for Sendai virus epitopes are induced by primary Sendai virus infection. We have previously shown that the acute effector CTL response to Sendai virus infection in C57BL/6 mice is directed predominantly against a single NP-derived epitope, NP_324-332/K^b (9). However, memory CTL precursors (CTLp) specific for both the dominant NP_324-332/K^b epitope and a subdominant epitope involving the same peptide, NP_324-332/D^b, are established following resolution of the primary infection (8). Interestingly, LDA indicates that CTLp specific for these two epitopes are induced at similar frequencies (approximately 0.05% CD8⁺ T cells) despite their disparate contributions to the acute phase of the response. To further investigate the induction of CD8⁺ T-cell memory for Sendai virus infection, we generated Sendai virus NP_324-332/K^b and NP_324-332/D^b tetramers to identify T cells with these antigen specificities directly. The specificities of these reagents were confirmed by using panels of T-cell hybridomas, and the tetramers were also shown to specifically stain T-cell lines or BAL fluid from acutely infected mice (16) (data not shown). Control tetramers folded with either influenza virus NP_366-374/D^b or MHV-68 p79/K^b did not stain T cells specific for these Sendai virus epitopes and have been described elsewhere (16, 34). Spleen cells were isolated from mice at various times after infection with Sendai virus and stained with Sendai virus-specific and control MHC tetramers plus CD8 and the activation markers CD44 and CD62L. As a control for these experiments, we also analyzed mice that had been infected with influenza virus. An unexpectedly high proportion (4 to 6%) of CD8⁺ T cells from Sendai virus memory mice (but not control influenza virus memory mice) stained positive with the NP_324-332/K^b tetramer, which detects T cells specific for the dominant epitope, at days 19 and 36 postinfection (Table 1). The memory NP_324-332/K^b-specific T-cell frequency was reduced at 89 days postinfection but still relatively high at 2.1% of the total CD8⁺ T cells. Despite the consistently high frequencies of memory CD8⁺ T cells specific for the dominant epitope, detection of T cells specific for subdominant NP_324-332/D^b was more variable and was, at most, one-half of that observed for NP_324-332/K^b (2.2% versus 4% on day 36 postinfection; Table 1). This pattern was much more pronounced in the effector T-cell population isolated directly from the lungs of acutely infected mice, where 70.7% of the CD8⁺ T cells stained with the NP_324-332/K^b tetramer and 1.1% of the CD8⁺ T cells stained with the NP_324-332/D^b tetramer (Table 1) (9).

Phenotype of memory K^b/NP_324-332-specific CD8⁺ T cells. We next investigated the expression of the activation-memory markers CD44 and CD62L on CD8⁺ T cells specific for dominant NP_324-332/K^b. All NP_324-332/K^b-specific cells expressed high levels of CD44 (Fig. 1). Expression of CD62L was more heterogeneous, with the majority of NP_324-332/K^b-specific cells being CD62L^lo and approximately 20% being CD62L^hi. Consistent with this finding, we noted a marked enrichment for NP_324-332/K^b-specific T cells among the CD8⁺ T cells with either a CD44^hi or a CD62L^lo phenotype (Table 1). For example, at day 36 postinfection, 4% of CD8⁺ T cells were NP_324-332/K^b tetramer positive, compared with 28.7% of CD8⁺/CD62L^lo cells. Interestingly, we did not see a similar enrichment of T cells specific for the subdominant NP_324-332/D^b epitope among CD8⁺/CD44^hi or CD8⁺/CD62L^lo T cells (Table 1). The reason for this difference is unclear but may be related to the subdominant status of this epitope.

View larger version (22K): [in this window] [in a new window]   FIG. 1.   Phenotype of NP324-332/Kb-specific CD8+ cells during memory. Spleen cells from B6 mice at 36 days postinfection with Sendai virus were stained with anti-CD8, anti-CD44, and anti-CD62L antibodies plus tetrameric NP324-332/Kb. The upper graph shows CD8 and NP324-332/Kb staining with a lymphocyte gate, and the lower two graphs show activation marker expression gated only on NP324-332/Kb-specific CD8+ cells. The data are representative of three experiments.

View larger version (36K): [in this window] [in a new window]   FIG. 2.   Presence of blasted NP324-332/Kb-specific CD8+ cells in the memory response. Spleen cells from C57BL/6 mice at various times (days [d]) after infection with Sendai virus were stained with anti-CD8 antibody plus tetrameric NP324-332/Kb. The graphs show forward scatter (FSC) and tetrameric NP324-332/Kb staining gated on all CD8+ cells. Shown are the percentages of CD8+ NP324-332/Kb+ cells in the quadrants.

Proliferative potential of memory NP_324-332/K^b-specific CD8⁺ T cells. Having established that the memory pool is heterogeneous with respect to activation-memory marker expression and cell size, we used an in vitro system to determine whether all of the cells in the memory pool had the same proliferative potential. In these and all subsequent experiments, mice were considered to be memory mice when they had recovered from an infection given at least 30 days previously. We took the approach of labeling cells with the fluorescent dye CFSE, which becomes progressively diluted out with each cell division (28). Spleen cells from Sendai virus-infected memory mice were removed, labeled with CFSE, and then restimulated in vitro with NP_324-332 peptide and IL-2. The culture was then sampled each day and stained with an anti-CD8 antibody and the NP_324-332/K^b tetramer, allowing us to monitor the division of NP_324-332/K^b-specific cells in vitro. As shown in Fig. 3, the outgrowth of NP_324-332/K^b-specific cells was easily detectable. In the first 2 days, there was little division of the tetramer-positive cells; however, these cells divided rapidly during the next 2 days and eventually expanded to account for approximately 66% of the CD8⁺ T cells in the culture after 6 days. The NP_324-332/K^b-specific cells divided in synchrony, and we did not observe discrete populations of cells which had undergone differing numbers of cell divisions, unlike those observed in primary T-cell responses in vitro (26, 30). Interestingly, we observed a population of NP_324-332/K^b+CD8⁺ T cells that did not divide in vitro and remained CFSE^hi. This could be seen most clearly at day 4 poststimulation (Fig. 3). At later time points, this population was less clear. This probably was due not to the disappearance of these cells but rather to dilution by the dividing cells in the culture. Both dividing and nondividing cells were alive and had high forward scatter, indicative of a blasted phenotype (data not shown). One possible explanation for the undivided population is that they were doublets consisting of a divided (CFSE^lo), tetramer-positive cell bound to a nondivided (CFSE^hi), nonspecific cell. To address this, we used the bandwidth feature of the FACScan to gate out all doublets. The nondivided, tetramer-positive cell population was not affected when doublets were excluded (data not shown), indicating that the data cannot be explained in this way.

View larger version (35K): [in this window] [in a new window]   FIG. 3.   Proliferation of NP324-332/Kb-specific CD8+ T cells after in vitro restimulation with the NP324-332 peptide. Spleen cells from Sendai virus-infected memory mice were labeled with the fluorescent dye CFSE and then placed in culture with the NP324-332 peptide plus IL-2. Each day, a fraction of the culture was harvested and stained with anti-CD8 antibody plus either tetrameric NP324-332/Kb (left) or anti-T-cell receptor antibody (right). The graphs display data obtained by gating on all live, CD8+ cells. The data are representative of three experiments.

View larger version (18K): [in this window] [in a new window]   FIG. 4.   Enumeration of the nonproliferating fraction of NP324-332/Kb-specific CD8+ cells. Spleen cells from Sendai virus memory mice were restimulated in vitro with the NP324-332 peptide plus IL-2 in the presence of 10 µg of BrdU per ml plus 1 µg of FdUrd per ml. A control culture received no BrdU-FdrUrd. Cultures were harvested after 48 h and stained with anti-CD8 and anti-BrdU antibodies plus tetrameric NP324-332/Kb. Results show staining with anti-BrdU antibody gated on all CD8+ NP324-332/Kb+ lymphocytes. Similar data were obtained in two experiments.

Function and phenotype of nondivided T-cell population. Previous studies have shown that restimulation of memory spleen cells from Sendai virus-infected mice with the NP_324-332 peptide was a very effective method of generating effector CTL in vitro (8). We therefore wanted to test whether this effector CTL activity resided solely in the divided population or whether both the divided and nondivided populations gave rise to effector T cells. Thus, we stimulated CFSE-labeled memory cells with the NP_324-332 peptide and separated the nondividing CD8⁺ NP_324-332/K^b+ CFSE^hi population from the dividing CD8⁺ NP_324-332/K^b+ CFSE^lo population 4 days later by fluorescence-activated cell sorting. Each population was tested for CTL activity on peptide-pulsed L-K^b target cells. As shown in Fig. 5, both populations of cells were equally effective at lysing NP_324-332-loaded L-K^b target cells, indicating that the nondivided cells were fully functional effector CTL after in vitro restimulation. These data demonstrate that a significant fraction of antigen-specific memory CD8⁺ T cells are able to differentiate into effector CTL in vitro without cell division.

View larger version (23K): [in this window] [in a new window]   FIG. 5.   Both divided and nondivided NP324-332/Kb-specific CD8+ cells possess an effector function. Spleen cells from Sendai virus memory mice were labeled with CFSE and then restimulated in vitro for 4 days with the NP324-332 peptide plus IL-2. Cells were then stained with anti-CD8 antibody plus tetrameric NP324-332/Kb prior to sorting by flow cytometry. The sample was gated to include only CD8+ cells and then sorted into NP324-332/Kb+ CFSElo (divided) and NP324-332/Kb+ CFSEhi (nondivided) cells. Sorted cells were then assayed for cytotoxicity in a 51Cr release assay using target cells pulsed with NP324-332 (closed symbols) or no peptide (open symbols). E:T ratio, effector-to-target cell ratio.

View larger version (33K): [in this window] [in a new window]   FIG. 6.   Phenotype of divided and nondivided NP324-332/Kb-specific CD8+ cells. Spleen cells from Sendai virus memory mice were labeled with CFSE and then restimulated in vitro for 4 days with the NP324-332 peptide plus IL-2. The culture was then stained with anti-CD8 antibody, tetrameric NP324-332/Kb, and antibodies recognizing lymphocyte activation markers. Gates were set to include live lymphocytes plus either CD8+ NP324-332/Kb cells (left) or CD8+ NP324-332/Kb+ cells (right). The data are representative of two experiments.

Origin of nondividing cells. To establish which cells from the memory pool gave rise to the nondivided CD62L^lo cells that we observed in vitro, we sorted different cell populations from the spleens of memory mice before CFSE labeling and in vitro restimulation with peptide. We sorted on the basis of CD44 and CD62L expression by using flow cytometry. Three cell populations were collected: CD44^hi CD62L^hi, CD44^hi CD62L^lo, and CD44^lo CD62L^hi. There was no population corresponding to the other combination of markers (CD44^lo CD62L^lo). These cells were labeled with CFSE; mixed with unlabeled, uninfected splenocytes plus the NP_324-332 peptide and IL-2; and then cultured for 4 days in vitro. Each culture was then stained with an anti-CD8 antibody and the NP_324-332/K^b tetramer. As shown in Fig. 7, there was a marked difference in the ratio of divided-to-undivided cells in the various cell fractions. In the unsorted sample, the ratio of undivided-to-divided cells was 1:14.5 whereas the CD44^hi CD62L^hi fraction gave a ratio of 1:2.5, demonstrating an enrichment for cells that did not divide in vitro. Conversely, the CD44^hi CD62L^lo fraction gave an undivided-to-divided ratio of 1:18, indicating enrichment for cells that divide in vitro in response to peptide. The CD44^lo CD62L^hi fraction did not contain a significant number of NP_324-332/K^b-specific cells, as expected from previous results (Fig. 1). There was no exact correlation between CD62L expression and proliferative potential, since both divided and nondivided populations were present in cultures of CD62L^hi and CD62L^lo cells. Nevertheless, these data clearly show that the CD62L^hi cells preferentially gave rise to cells with low proliferative potential whereas CD62L^lo cells gave rise to cells with high proliferative potential.

View larger version (27K): [in this window] [in a new window]   FIG. 7.   Proliferative potential segregates with the CD62L phenotype. Spleen cells from Sendai virus memory mice were stained with anti-CD44 and anti-CD62L antibodies and then sorted by flow cytometry into the three populations shown. Cells were labeled with CFSE and then added to an equal number of spleen cells from uninfected mice plus the NP324-332 peptide and IL-2 and cultured for 4 days. The cultures were then harvested and stained with anti-CD8 antibody and tetrameric NP324-332/Kb. The graphs show CFSE fluorescence and NP324-332/Kb staining gated on live CD8+ lymphocytes. Similar data were obtained in two experiments.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The data presented here show that memory CD8⁺ T cells specific for the dominant epitope are maintained at elevated levels for long periods after Sendai virus infection. This population was heterogeneous with respect to cell size and expression of the activation-memory marker CD62L. In addition, CD62L expression correlated with the ability of NP_324-332/K^b-specific CD8⁺ T cells to proliferate in vitro in response to antigen. This demonstrates that the memory CD8⁺ T-cell pool consists of subpopulations with different functional capabilities that are associated with expression of CD62L.

The high frequency of memory CD8⁺ T cells induced by Sendai virus infection (2 to 6% of CD8⁺ spleen cells) was surprising given that influenza virus, which induces a similar infection of the lung, only generates memory CD8⁺ T cells on the order of 0.5 to 0.8% of CD8⁺ spleen cells (16). In this regard, Sendai virus is more like LCMV, which induces antigen-specific memory cells up to 8% of the CD8⁺ T cells in the spleen (6, 29). Importantly, these data demonstrate that not all respiratory infections generate low frequencies of memory CD8⁺ T cells and suggest that LCMV and influenza virus may merely represent two ends of the spectrum. It is unclear what controls differences in magnitude between the memory CD8⁺ T-cell pools induced by infections with different viruses. However, it has been suggested that the differences may be related to antigen load in lymphoid tissues, since LCMV directly infects the spleen whereas influenza virus only infects the lung. In support of this idea, Gallimore et al. have shown that the size of the immune response is proportional to the virus load in the LCMV system (17). However, the efficiency with which Sendai virus induces CD8⁺ memory cells suggests that this is not the primary factor. Like influenza virus, Sendai virus replicates only in the lung and generates similar total viral loads. Live virus is not found in the spleen, and although antigen-presenting cells can migrate from the lung to the spleen, they are present at very low frequencies (39). It is possible that the amount of antigen carried to lymphoid tissue is larger in Sendai virus infection than in influenza virus infection; however, there is little information about the influenza virus system so a direct comparison is difficult. Interestingly, the acute effector responses to these two viruses are substantially different. Sendai virus infection induces a small inflammatory response in the lung, compared to the strong inflammatory response induced by influenza virus. However, the majority (70%) of CD8⁺ T cells in the lungs of Sendai virus-infected mice are specific for the dominant epitope. This contrasts strongly with the relatively small fraction (15%) of dominant-epitope-specific CD8⁺ T cells in influenza virus-infected mice. Even in a secondary immune response to influenza virus infection, only half of the CD8⁺ T cells in the lungs are demonstrably specific for the dominant epitope (16). Thus, it is possible that the difference between the frequencies of antigen-specific memory cells established in these two systems simply reflects the CD8⁺ T-cell composition of the prior acute response in the lung.

We found previously that the frequencies of memory CD8⁺ T cells specific for the dominant NP_324-332/K^b and subdominant NP_324-332/D^b epitopes were similar by using LDA (on the order of 1:2,000 CD8⁺ T cells) (8). In contrast, more direct analysis of the dominance hierarchy among memory T cells using MHC tetramers reveals a pattern of immunodominance that mirrors that of the primary response. A lower frequency of subdominant-epitope-specific CD8⁺ T cells has also been reported previously in listeria infection (5). This difference may relate to the ability of these cells to survive and proliferate in vitro, a necessary requirement for detection by LDA. We could clearly detect the proliferation of NP_324-332/D^b-specific cells in vitro by using CFSE (data not shown); however, due to the small and variable population size it was not possible to make a meaningful comparison with the NP_324-332/K^b response.

The stimuli that perpetuate high frequencies of memory CD8⁺ T cells are unknown. A role for persistent antigen cannot be ruled out (18), although several reports have argued against this idea (1, 21, 25). It has recently emerged that activated-memory cells (CD44^hi CD62L^lo CD45RB^hi) turn over rapidly in vivo (36, 37), and this can be enhanced by the administration of type 1 interferon or IL-15 (36, 42). Sendai virus is a potent inducer of type 1 interferon; however, production presumably ceases in the lung after the infection is cleared. What triggers the synthesis of type 1 interferon or IL-15 during memory is unclear. We attempted to increase the pool of virus-specific memory CD8⁺ T cells by administering IL-15 by the protocol used by Zhang et al. (42); however, we detected no change (data not shown). It is likely that this cytokine affects cell turnover and homeostasis (27) rather than the absolute size of the memory pool.

We took advantage of the relatively high memory T-cell frequency to investigate functional differences within this population. Our experimental system is similar to that reported recently by Oehen and Brduscha-Riem (30), who studied the in vitro response of T-cell receptor transgenic T cells to an antigen from LCMV. These investigators showed that the memory CTL response consisted of two distinct populations: a CD62L^lo population which could exert a rapid effector function and a CD62L^hi population which required prolonged contact with target cells before developing an effector function. Our data agree with this study and take it an important step further by extending the model to a natural pathogen infection of wild-type mice. We also show a functional segregation based on CD62L expression during memory. Our study demonstrates that the proliferative potential of antigen-specific CD62L^lo cells is greater than that of CD62L^hi cells. There is no absolute correlation between proliferative potential and CD62L expression, since we observed a population of cells derived from CD62L^lo cells which failed to proliferate and a population of cells from CD62L^hi cells which did proliferate. It will be interesting in future studies to elucidate whether these populations with altered proliferative potentials modulate CD62L expression accordingly.

The relationship between the proliferative potential of memory CD8⁺ T cells and their effector function is unclear. It is clear that Sendai virus-specific memory CD8⁺ T cells are capable of rapid antigen responsiveness, as the frequency of cells which produce gamma interferon in response to 5 h of antigen stimulation is similar to that observed by using tetramer staining (39a). This is also true of other virus infections (16, 29). Other investigators have reported that the persistence of activated CD8⁺ T-cell memory is necessary for protection from peripheral virus challenge (4, 24). In contrast, resting memory cells are sufficient to protect from systemic viral challenge but appear to lack the ability to recirculate into peripheral tissues. Further evidence for the persistence of effector cells is contained in the recent report of Selin and Welsh (32), who used sensitive target cells to show that cytolytically active CTL could be detected in the spleen for the lifetime of the mouse after LCMV challenge. These effector cells may be needed to eliminate virus rapidly from peripheral sites, whereas another population of memory cells takes longer to differentiate into effector cells and develops into a "second wave" of effector cells (4, 24). These latter cells may then be of importance in preventing systemic spread of the virus if the effector memory population fails to clear the virus. Our data showing that some memory T cells proliferate more than others lead to the speculation that these "second-wave" cells may be those that can proliferate and thus clonally expand to combat an infection. In contrast, the cells with low proliferative potential may comprise the first wave of effector cells which are able to attack infected cells more quickly.

In conclusion, we used Sendai virus infection as a model system with which to address some of the issues concerning CD8⁺ T-cell memory that had been demonstrated in other experimental systems. We showed that the memory CD8⁺ T-cell population consists not of one homogeneous population but rather of subpopulations of cells with different phenotypes and capacities to proliferate in response to antigen. This has profound implications in terms of our understanding of immunological memory because it indicates that different memory CD8⁺ T-cell populations perform different functions in vivo. A fuller understanding of how these different functions arise may lead to improved methodologies for inducing protective immune responses.