|Home | About | Journals | Submit | Contact Us | Français|
The CD8+ T cell response to infection is characterized by the appearance of short-lived (CD127low killer cell lectin-like receptor G 1–high) and memory-precursor (CD127high killer cell lectin-like receptor G 1–low) effector cells. How and when central-memory T (TCM; CD62Lhigh CCR7+) cell and effector-memory T(TEM; CD62Llow CCR7−) cell subsets are established remains unclear. We now show that the TCM cell lineage represents an early developmental branchpoint during the CD8+ T cell response to infection. Central-memory CD8+ T cells could be identified prior to the peak of the CD8+ T cell response and were enriched in lymphoid organs. Moreover, the kinetics and magnitude of TCM cell development were dependent on the infectious agent. Furthermore, the extent of early Ag availability, which regulated programmed death-1 and CD25 expression levels, controlled the TCM/TEM cell lineage decision ultimately through IL-2 and IL-15 signaling levels. These observations identify key early signals that help establish the TCM/TEM cell dichotomy and provide the means to manipulate memory lineage choices.
In recent years, much has been elucidated regarding how CD8+ T cell responses unfold. Postinfection, extremely rare naive Ag-specific CD8+ T cells (1–3) encounter an APC undergoing an exquisitely orchestrated and somewhat prolonged activation phase (4). Following their activation, Ag-specific CD8+ T cells undergo a rapid expansion, after which only ~5–10% of the pathogen-specific CD8+ T cells are maintained as a memory population. The resulting memory population generally contains increased numbers of Ag-specific CD8+ T cells, compared with the naive pool. These cells exhibit altered homing patterns, increased TCR avidity, and enhanced cytokine production, all of which enable them to respond with increased vigor and potency to future encounters with the same pathogen (5–8).
Questions remain about how, when, and where the decision is made for activated CD8+ T cells to develop into memory cells. A recent study demonstrated that all effector and memory cell populations generated following Listeria monocytogenes infection could be generated from a single naive Ag-specific CD8+ T cell (9). Intriguingly, it has been elegantly demonstrated that asymmetric cell division as early as the first cell division can result in proximal and distal daughter CD8+ T cells that have different lineage fates (10). These, and other studies (11), clearly indicate that CD8+ T cells with memory potential are found in the effector CTL pool. Indeed, at the peak of the CD8+ T cell response a small subset of effector cells retain CD127 (IL-7Rα) expression, and these cells go on to form the long-lived memory pool (12, 13). Further analyses have demonstrated that the effector CD8+ T cell population contains both a small population of memory-precursor effector cells (MPECs) and a larger number of terminally differentiated short-lived effector cells (SLECs), distinguished on the basis of CD127 and killer cell lectin-like receptor G1 (KLRG1) expression (14, 15).
Additional layers of complexity exist within the memory CD8+ T cell population with respect to phenotype, function, and anatomic location. For example, memory CD8+ T cells are heterogeneous with respect to homing molecule expression and contain at least two distinct populations: CD62Lhigh CCR7+ central-memory T (TCM) cells and CD62Llow CCR7− effector-memory T (TEM) cells (16, 17). TCM cells are predominantly found within secondary lymphoid organs, as well as the blood and spleen. In contrast, TEM cells are primarily found within peripheral tissues (i.e., lung, gut, and liver), as well as the blood and spleen (18, 19). It is thought that the TEM cell population provides immediate protection at environmental barriers, whereas the TCM cell population provides a second layer of protection upon Ag rechallenge (20). Which memory subset plays a role in mounting a secondary response is dependent, in part, on the location of Ag challenge and characteristics of the pathogen (21–24). Thus, understanding the salient features of memory T cell subsets requires consideration of the parameters of each type of infection or immunization route.
It has been known for some time that CD8+ memory T cell populations gradually shift from being largely CD62Llow to primarily CD62Lhigh (25). Two competing hypotheses have been proposed to explain this phenomenon. In the first model, the TEM cell population is largely transient in nature and gives rise to the TCM cell population (21, 26); thus far, this effect appears to be the result of abnormally high precursor frequencies used in adoptive transfer systems (1, 21, 27–29). In the opposing model, the TEM and TCM cell pools, at least based on CD62L expression, are independent lineages with transition of the memory population over time to a predominantly TCM cell phenotype due to the higher homeostatic proliferative rate of TCM cells (1, 27). Although previous studies have identified memory precursors present at the peak of the response, the precise origin of the TEM and TCM cell subsets has yet to be elucidated. In this paper, we examined early effector cell (EEC) differentiation events after L. monocytogenes and vesicular stomatitis virus (VSV) infection. These studies identify the early origin of the TCM cell population and elucidate the signals required for the development of this subset. These findings appreciably enhance our knowledge regarding the temporal and physical interactions regulating memory CD8+ T cell generation.
Female C57BL/6 and B6-Ly5.2 mice between 5 and 8 wk old were purchased from the National Cancer Institute. Female B6.129S4-Il2ratm1Dw/J (CD25−/−) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) or were a kind gift from Dr. Charles Surh (Scripps Institute). C57BL/6 IL-15−/− mice (30) were bred in the University of Connecticut Health Center animal facility. All animal protocols were approved by the University of Connecticut Health Center Animal Care Committee.
Bone marrow cells were obtained from femurs and tibias of B6.129S4-Il2ratm1Dw/J and B6-Ly5.2 mice. Recipient mice were irradiated with ~1000 rads and subsequently injected i.v. with 106 bone marrow cells (2:1 ratio of CD25−/−/B6). Chimeras were rested 6–8 wk before use in experiments.
Both the rVSV expressing OVA (31) and recombinant L. monocytogenes expressing OVA (32) have been previously described. Mice were infected i.v. with either 105 PFUs of VSV-OVA or 103 CFUs of L. monocytogenes-OVA (LM-OVA).
Single-cell suspensions were prepared by collagenase digestion, as previously described (18). The H-2Kb tetramer containing either the OVA-derived peptide SIINFEKL or VSV nucleoprotein (VSV-N)-derived peptide RGYVYQGL was generated as previously described (33). Analysis of the Ag-specific CD8+ T cells early postinfection by tetramer enrichment has already been described (1). For general staining, 107 lymphocytes per milliliter were incubated with the appropriate peptide: MHC class I tetramers, anti-CD8a (53-6.7; BioLegend, San Diego, CA), and Fc block (2.4G2; BD Pharmingen, San Diego, CA) for 1 h at room temperature. Cells were then washed and stained with anti-CD62L (MEL-14; eBioscience, San Diego, CA), anti-KLRG1 (2F1; Abcam, Cambridge, U.K.), anti-CD127 (A7R34; eBioscience), anti-CD44 (IM7; BioLegend), anti-CD25 (PC-61; BioLegend), anti-PD1 (RPM1-30; BioLegend), and anti-CD11a (2D7; BD Biosciences, San Jose, CA) for 30 min at 4°C. Samples were analyzed on an LSR-II (BD Biosciences), and data analysis was accomplished using FlowJo (Tree Star, Ashland, OR).
BrdU was administered to infected mice in their drinking water (0.8 mg/ml). Spleen cells were then cell surface stained as above with SIINFEKL/Kb tetramer, anti-CD8a, anti-CD62L, and anti-CD11a. After this, cells were stained with anti-BrdU according to the BrdU flow kit protocol (BD Biosciences).
The mAb specific for the OVA derived SIINFEKL peptide presented in the context of H-2Kb (25-D1.16) has been previously described (34). The 25-D1.16 mAb and an isotype-matched irrelevant control Ab (MOPC-21) were purchased from the National Cell Culture Center, and mice were injected i.p. with graded amounts of the 25-D1.16 or control mAb at the indicated times. CD62L expression was quantified at day 7 or day 42 postinfection by flow cytometry, as described above.
Statistical significance was determined by either a Student t test or ANOVA, using Prism 5 (Graphpad Software). Significance was set at p < 0.05.
Both VSV and L. monocytogenes have been proposed as vaccine vectors, and secondary responses against these pathogens are mediated by distinct memory subsets (23). Given this, we examined CD62L expression after VSV or L. monocytogenes infection. At the peak of the CD8+ T cell response, only a few CD62Lhigh cells were detectable in the spleen (Fig. 1A). Over time, CD62Lhigh cells steadily increased within the OVA/Kb-specific CD8+ T cell population (Fig. 1A). Interestingly, OVA/Kb-specific CD8+ T cells induced by L. monocytogenes infection had a significantly higher proportion of CD62Lhigh cells when compared with VSV infection at early memory time points. However, by ~125 d postinfection, the frequency of CD62Lhigh cells was similar for both VSV and L. monocytogenes infections in the spleen. When the lungs were examined, a similar trend was observed, although the transition toward CD62Lhigh cells occurred more slowly (Fig. 1A). Within the lymph nodes, emergence of the CD62Lhigh population was rapid. By day 21 postinfection, most L. monocytogenes-specific cells were CD62Lhigh, whereas a proportion (~25%) of VSV-specific cells lacked CD62L (Fig. 1A). Thus, L. monocytogenes infection, compared with VSV infection, drove more rapid generation of CD62Lhigh central memory cells.
Previous work has indicated that the memory CD8+ T cell population transitions from a CD62Llow to CD62Lhigh phenotype over time (25). This phenomenon is due to the increased turnover rate of the CD62Lhigh relative to the CD62Llow memory cells (1, 21, 27). As we have just shown, the kinetics of the transition of the memory population from CD62Llow to CD62Lhigh differed following VSV and L. monocytogenes infection. To determine the mechanism underlying this dichotomy, we compared memory subset proliferation after each infection. VSV-OVA or LM-OVA memory mice were treated with BrdU in their drinking water for 4 wk, and BrdU incorporation into the CD62Lhigh and CD62Llow OVA/Kb-specific CD8+ T cells was quantified. Following either infection, ~40% of the CD62Lhigh memory CD8+ T cells had incorporated BrdU (Fig. 1B) and, in both cases, CD62Llow memory cells incorporated less BrdU than did CD62Lhigh cells. However, a significantly greater fraction of the CD62Llow cells in VSV-primed mice incorporated BrdU, compared with the same subset in L. monocytogenes-infected mice (~28% versus ~15%; p < 0.01). A greater ratio of CD62Lhigh memory cells incorporating BrdU was observed following L. monocytogenes infection versus VSV infection (Fig. 1B), which likely accounts for the delayed transition to a CD62Lhigh memory population following VSV infection. Furthermore, enhanced proliferation of the VSV-specific TEM cell population is likely due to the presence of low-level persistent Ag after VSV infection (35).
The origin of the CD62Lhigh memory CD8+ T cell population remains controversial. Originally, it was postulated that CD62Llow cells were capable of re-expressing CD62L, resulting in the gradual generation of the CD62Lhigh memory population (21, 26). However, other studies suggest that the CD62Lhigh and CD62Llow memory populations are distinct lineages that are not capable of interconverting (1, 27). As just discussed, the conversion at the population level of the memory pool toward increased CD62L expression is likely the result of differences in turnover rates between the subsets (1, 21, 27). Because the effector CD8+ T cell population can be subdivided into SLEC and MPEC populations, which are CD127low KLRG1high and CD127high KLRG1low, respectively (14, 15), we hypothesized that a proportion of the MPEC population might retain expression of CD62L early postinfection. To test this hypothesis, CD62L expression was quantified on each of the detectable effector cell populations after VSV or L. monocytogenes infection. On day 7 postinfection, the MPEC population in the spleen and lymph nodes contained a readily identifiable population of CD62Lhigh cells (Fig. 2A). In contrast, both the SLEC and EEC, which is CD127low KLRG1low in phenotype, populations of OVA/Kb-specific CD8+ T cells in the spleen, lymph nodes, or lungs largely lacked CD62L expression (Fig. 2A). The EEC population from the lymph nodes did contain a small population of CD62Lhigh cells, but these were extremely few in numbers and may represent cells in transition to the MPEC population. Notably, a greater proportion of the MPECs present after L. monocytogenes infection retained CD62L expression (Fig. 2A). In addition, CD62L expression levels of OVA/Kb-specific CD8+ T cells were tissue specific. Few CD62Lhigh CD8+ T cells could be detected in the lung parenchyma, but a much higher frequency of CD62Lhigh CD8+ T cells was observed in the lymph nodes (Fig. 2A).
When the kinetics of CD62Lhigh MPEC development in the spleen was examined over time, an interesting pattern emerged. On day 5 postinfection, the frequency of CD62Lhigh cells was similar between VSV- and L. monocytogenes-infected mice (Fig. 2B). Over the next 5 d, the proportion of CD62Lhigh cells declined following both infections, but declined significantly less in mice infected with L. monocytogenes (Fig. 2B). Over the next ~4 mo, CD62Lhigh cells increased in the MPEC population at a much greater rate with L. monocytogenes infection than with VSV infection, likely as a result of two factors: 1) an early increase in the frequency of CD62Lhigh cells within the MPEC population that is then maintained (Fig. 2A) and 2) the proliferative differences in the TEM cell populations described above (Fig. 1B). Overall, these results shed light on the relationship of the development of the effector cell subsets and the emergence of the CD62Lhigh memory population.
The factors that regulate the generation of the CD62Lhigh and CD62Llow memory populations are not well defined. In vitro priming studies suggest that weak stimulation may preferentially generate cells of a TCM cell phenotype, whereas prolonged stimulation will generate TEM phenotype cells (36, 37). In vivo, adoptive transfer of graded numbers of TCR transgenic (Tg) CD8+ T cells suggests that increased competition leads to generation of higher numbers of CD62Lhigh cells (21, 27, 29, 38). However, adoptive transfer of high numbers of TCR Tg cells results in a more rapid clearance of L. monocytogenes from the spleen and reduced levels of inflammatory cytokines (29, 39). Thus, these in vivo studies cannot distinguish between the effects of Ag levels and the inflammatory milieu.
Therefore, we wished to probe whether the CD62Lhigh and CD62Llow populations received similar levels of TCR stimulation. Programmed death-1 (PD-1), although functioning as a negative regulator in most cases (40), is rapidly upregulated after TCR engagement (41) or γc cytokine signaling (42). We had previously noted that PD-1 expression declined with increased competition using graded numbers of OT-I TCR Tg cells (data not shown). Thus, PD-1 expression at the peak of the CD8+ T cell response appeared to be tunable to the overall strength of stimulation received by the responding cells, which is the sum of TCR, costimulatory, and cytokine signals. Because adoptive transfer of high numbers of TCR Tg cells resulted in weaker PD-1 upregulation and those cells tended to be CD62Lhigh, we next wanted to examine whether PD-1 expression and CD62L expression were related during endogenous CD8+ T cell responses. To test whether this held true in the endogenous CD8+ T cell population, C57BL/6 mice were infected with either VSV-OVA or LM-OVA. At the peak of the OVA/Kb-specific CD8+ T cell response, PD-1 expression was measured on the CD62Lhigh and CD62Llow subsets in the lymph nodes (Fig. 3A) and spleen (data not shown). Interestingly, PD-1 expression inversely correlated with CD62L expression. Thus, CD62Llow MPEC expressed PD-1, whereas CD62Lhigh MPECs largely lacked PD-1. These data suggested that theCD62Lhigh population had received a weaker overall activation stimulus than did the CD62Llow subset.
With the previous PD-1 data in mind, we wanted to directly test the role of TCR triggering and Ag levels on the outcome of endogenous OVA/Kb-specific CD8+ T cells without altering the clearance of the pathogen or inflammatory environment. To achieve this, mice were injected with either graded amounts of 25-D1.16 mAb, which is specific for the OVA-derived SIINFEKL peptide presented in the context of H-2Kb (34), or with an isotype-matched control Ab (MOPC-21). After injection, mice were infected with VSV-OVA, and 7 or 42 d later, OVA/Kb-specific CD8+ T cells in the spleen and lymph nodes were analyzed. In line with the previous observation, injection of increasing amounts of 25-D1.16 mAb resulted in a dose-dependent decrease in PD-1 expression on OVA/Kb-specific MPECs (Fig. 3B). Injection of increasing amounts of the 25-D1.16 mAb also decreased the magnitude of the OVA/Kb-specific CD8+ T cell response (Fig. 4A). Interestingly, 25-D1.16 mAb treatment did not affect the overall distribution of the EEC, MPEC, and SLEC subsets (data not shown). In contrast, CD62L expression was substantially altered by the injection of the 25-D1.16 mAb (Fig. 4A). Treatment with 250 μg of the 25-D1.16 Ab, but not 50 μg, resulted in a significantly higher proportion of the Ag-specific MPECs expressing CD62L (p < 0.001) (Fig. 4A). Furthermore, this difference in CD62L expression was maintained into memory (Fig. 4C). Similar data were obtained using the 25-D1.16 mAb blockade during LM-OVA infection (Fig. 4D). As a control, we also examined the CD8+ T cell response against the VSV-N in the VSV-OVA infected animals and found no differences in the magnitude of the VSV-N/Kb-specific CD8+ T cell response or in the phenotype of the responding cells (Fig. 4B). This finding indicated that the 25-D1.16 mAb did not deplete APC or affect “bystander” responses. Thus, competition for Ag and apparent TCR signal strength regulated not only CD8+ T cell expansion but also TCM cell development.
The timing of naive cell entry into the response has also been proposed to regulate memory development (43). Using adoptive transfer systems, cells added into an ongoing response tend to preferentially form TCM-type memory cells (29, 44, 45). In addition, the duration of T cell–APC interaction required to drive memory development has been a matter of discussion. Early studies suggested that only a few hours of stimulation with cognate Ag were needed to drive a productive response (46, 47), although recent work suggests that a more prolonged period (72–96 h) of Ag availability is required for an optimal T cell response (4, 15, 48, 49). With this in mind, we asked when the CD62L expression pattern was determined. To achieve this, mice were infected with VSV-OVA and treated with 250 μg of the 25-D1.16 mAb at various times. At 7 d postinfection, the OVA/Kb-specific CD8+ T cell response was monitored in the spleen and lymph nodes. Interestingly, effector cell expansion and CD62L expression were differentially regulated. Mice treated with 25-D1.16 mAb just prior to infection had a significantly reduced OVA/Kb-specific CD8+ T cell response, with a greater proportion of the responding OVA/Kb-specific CD8+ T cells being CD62Lhigh (Fig. 5A). Whereas 25-D1.16 mAb blockade as late as 72 h postinfection impaired expansion, mAb-induced modulation of CD62L expression occurred only up to 48 h postinfection and was most notable when mAb was given no later than 24 h postinfection (Fig. 5A). Similarly, during LM-OVA infection, maximal Ag availability for up to 48 h was necessary to induce changes in CD62L regulation, whereas 96 h was needed for optimal expansion (Fig. 5B). Using total numbers of OVA/Kb-specific CD8+ T cells, we also calculated the TEM/TCM cell ratio among MPECs after VSV infection and 25-D1.16 mAb treatment on different days (Fig. 5C). We observed that not only was the overall number of MPECs inhibited by 25-D1.16 treatment early, but also the ratio of TEM/TCM phenotype cells was again skewed toward TCM cells (Fig. 5C). The ratio in control mice was 9.6:1, whereas treatment at day 0 or day 1 decreased the ratio to 2.8:1 and 2:1, respectively. Treatment on days 2 or 3 altered the ratio to ~4:1, but the effect was waning with day 4 treatment (7:1). The magnitude of the effect correlated with the extent of the overall inhibition of the response resulting from Ag blockade. These data demonstrated that the concentration of available Ag during the first 3 d of CD8+ T cell priming was important in regulating CD62L expression, whereas Ag accessibility was necessary for up to 96 h for optimal CD8+ T cell expansion, as previously suggested (4). These data indicated that the TCM/TEM cell lineage choice is made earlier than previously described based on the “latecomer” hypothesis (29, 44, 45), but does not rule out the possibility of such a phenomenon occurring.
In addition to TCR-mediated signals, cytokines of the γc family are also important in memory T cell development and survival (50). Moreover, in vitro studies have demonstrated that IL-2 and IL-15 can generate effector CD8+ T cell populations that resemble TEM and TCM cell populations, respectively (51, 52). It was therefore of interest to determine the roles of these cytokines in memory subset differentiation. CD25 expression by Ag-specific CD8+ T cells was maximal at day 4 postinfection (Fig. 6A). By comparison, alterations of CD62L on the Ag-specific CD8+ T cells following 25-D1.16 mAb administration only occurred when the mAb was administered prior to maximal CD25 expression (Figs. 4, ,5).5). Thus, we tested whether 25-D1.16 mAb administration could alter CD25 levels on the responding Ag-specific CD8+ T cells. Indeed, 4 d after LM-OVA infection, Ag-specific CD8+ T cells expressed high levels of CD25 in the control mice, but cells from mice treated with 250 μg of 25-D1.16 had substantially lower CD25 levels (Fig. 6B). Thus, the effect of Ag competition could be mediated downstream by cytokines of the γc family.
The receptors for both IL-2 and IL-15 share two common receptor subunits, CD122 (IL-2/15rβ) and CD132 (γc receptor), and each has unique α-chains to form the high-affinity receptor (50). To test the in vivo role of IL-2 and IL-15 in the generation of the TCM and TEM memory cell subsets, C57BL/6:CD25−/− mixed bone marrow chimeras, IL-15−/− mice, and C57BL/6 mice were infected with LM-OVA. At the peak of the CD8+ T cell response (day 9), splenic OVA/Kb-specific MPECs were analyzed for CD62L expression. At this time, CD25−/− OVA/Kb-specific CD8+ T cells had a significantly increased frequency of CD62Lhigh cells (p = 0.0325) (Fig. 6C), whereas in the absence of IL-15, the frequency of CD62Lhigh cells was significantly decreased (p = 0.0179) (Fig. 6C). These data extend the prior in vitro studies (51, 52) and demonstrated that in vivo IL-2–derived signals promote the downregulation of CD62L and formation of the TEM cell population, whereas IL-15–derived signals promote the expression of CD62L and formation of TCM cells. Because the 25-D1.16 blockade inhibited expression of CD25 on the responding CD8 T cells, we next asked whether 25-D1.16 blockade of Ag presentation and CD25 expression had overlapping functions in regulating CD62L expression. For these experiments C57BL/6:CD25−/− mixed bone marrow chimeras were treated with 250 μg of 25-D1.16 or MOPC-21. The mice were then infected with 103 CFU of LM-OVA, and the generation of CD62Lhigh TCM cells within the MPEC population on day 9 was analyzed. C57BL/6:CD25−/− mixed bone marrow chimeras treated with MOPC-21 had a low frequency of CD62Lhigh TCM cells in the C57BL/6 compartment and an elevated frequency in the CD25−/− compartment. Interestingly, when the C57BL/6:CD25−/− mixed bone marrow chimeras were treated with 25-D1.16, there was an enhancement of CD62Lhigh TCM cells only of C57BL/6 origin, but not of CD25−/− origin (Fig. 6D). Taken together, these data demonstrated that CD62Lhigh TCM cells are generated when Ag is limiting, and this occurs through the limitation of IL-2–mediated signals.
Because immunological memory is the foundation of vaccination, an understanding of the factors regulating the development of the memory population is critical. Furthermore, it has been illustrated that TCM and TEM cells have different recall and protective abilities, depending on the challenge infection (21–24, 53). Therefore, understanding the factors governing and regulating the differentiation of the memory subsets is important for generating better vaccines. Strikingly, we have demonstrated that CD62Lhigh CD8+ T cells were mostly found within the MPEC population and could be identified as early as day 5 postinfection (Fig. 2). Furthermore, regulation of CD62L was tied to the overall strength of the activation signal received by the Ag-specific CD8+ T cell early during priming (Figs. 4–6). Our data indicate that the overall potency of the activation signal is, at least in part, the net result of integrating pMHC-TCR engagement and γc cytokine signals.
The origin of the memory CD8+ T cell population has long been debated. A recent report from Busch and colleagues (9) elegantly demonstrated that a single naive CD8+ T cell could give rise to all the different subsets of effector and memory CD8+ T cells. Thus, it is critical to determine the mechanism(s) by which effector cells survive to form the memory population. Furthermore, it is important to understand the relationship of the TCM/TEM cell dichotomy within the different effector CD8+ T cell populations. Earlier work demonstrated that a population of effector CD8+ T cells retains the expression of CD127 (IL-7Rα) (12, 13). However, IL-7 signals are not necessary for the survival of that population (54, 55). Recent work, using CD127 in combination with KLRG1 expression, has more extensively defined the effector cell populations present during many infections, whereby memory precursor effector cells are defined as CD127high KLRG1low and SLECs are defined as CD127low KLRG1high (14, 15, 56, 57). However, the relationship of the TCM/TEM cell dichotomy was not explored. In this paper, we demonstrate that only when the MPEC population became detectable did CD62Lhigh cells appear. Interestingly, over the next week the frequency of CD62Lhigh OVA/Kb-specific CD8+ T cells actually decreased before slowly shifting to a TCM cell phenotype, as previously described (25). Previous in vitro studies showed that CD62L expression is regulated in a three-step process whereby initial downregulation is mediated by proteolytic cleavage, followed by a rapid re-expression of CD62L and lastly a gradual genetic modulation of CD62L expression (58, 59). This three-step model supports our in vivo observations examining the early dynamics of CD62L expression on activated Ag-specific CD8+ T cells. 1) EECs lose cell surface expression of CD62L likely by proteolytic cleavage after initial CD8+ T cell activation; 2) as the immune response continues, heterogeneity within the MPEC population is generated by a small proportion of the Ag-specific CD8+ T cells that are genetically competent to re-express CD62L; and 3) the remainder of the MPECs, as well as EECs and SLECs, undergo epigenetic modification of the CD62L promoter region, prohibiting further gene expression. Thus, memory cell heterogeneity and trafficking patterns originated within the first week of infection. With time, the TCM cell population later dominates the memory pool due to its increased turnover rate, as previously seen (21, 27). However, as shown in our work, the rate of conversion differs between pathogens.
The identity of the precise signals that regulate CD62L expression and the differentiation of the TCM cell population in vivo have been unclear. Competition for Ag early in the priming of naive CD8+ T cells will alter the overall signal strength delivered to the responding CD8+ T cells. Previous studies have demonstrated that adoptive transfer of large numbers of TCR Tg CD8+ T cells leads to the generation of more TCM cells (1, 21, 27, 29). This effect is thought to be due to high competition for Ag, but the adoptive transfer of high numbers of TCR Tg CD8+ T cells can also alter the inflammatory milieu owing to rapid clearance of the infection (29, 39). Thus, we have used the 25-D1.16 (anti-SIINFEKL/Kb) mAb in a novel manner to specifically limit SIINFEKL/Kb Ag presentation in vivo. Using this method, we have observed two intriguing aspects of CD8+ T cell activation: expansion, and memory differentiation. First, optimal CD8+ T cell expansion required prolonged Ag availability Our previous finding indicates that late APC–T cell interactions form in an Ag-dependent manner (4), and our current data now show that these events enhance T cell expansion. In contrast, restriction of Ag availability only during the first 2–3 d postinfection altered CD62L expression. Therefore, a kinetic dichotomy exists between the requirement for Ag for optimal expansion and CD62L regulation.
One classic consequence of TCR-mediated signaling is the upregulation of CD25 (60). Our data demonstrated that CD25 expression peaked between days 3 and 5 (Fig. 6A) (57, 61, 62), a time just beyond when 25-D1.16 mAb administration became ineffective at modulating CD62L expression. Indeed, restricting Ag availability limited CD25 expression on the responding Ag-specific CD8+ T cells (Fig. 6B) and also decreased PD-1 expression (Fig. 3B). Furthermore, the observation that the strength of TCR signaling regulates CD25 expression is supported by the fact that vaccination with a weak altered-peptide ligand resulted in diminished CD25 expression on Ag-specific CD8+ T cells (63). In addition, at least in vitro, high levels of inflammation generated using unmethylated CpG DNA and/or IL-12 can enhance CD25 expression during T cell activation (64). Furthermore, limiting inflammation during L. monocytogenes infection by treatment of mice with ampicillin resulted in enhanced TCM cell formation, but the expression of CD25 was not explored (65, 66). Interestingly, IL-21 can limit the expression of CD25 on responding CD8+ T cells (67) and enhances CD62L expression (67, 68). These results fit well with our observation that IL-2 signaling results in decreased CD62L expression, whereas IL-15 signaling promotes CD62L expression in vivo. Furthermore, limiting Ag availability did not enhance CD62L expression in CD25-deficient cells, suggesting that the regulation of CD62L is ultimately controlled by the levels of IL-2 and IL-15 signaling.
In molecular terms, recent reports have demonstrated that the PI(3)K and mTOR signaling networks play a critical role in regulating T cell migration through control of CD62L, CCR7, and spingosine-1-phospate receptors (52, 69). Cantrell and colleagues (52, 70) have demonstrated that IL-2 strongly activates the PI(3)K pathway, leading to mTOR activation and the subsequent genetic silencing of CD62L, whereas IL-15 only weakly activates the PI(3)K and mTOR axis and results in the maintenance of CD62L expression. Furthermore, a recent report found that modulating mTOR activity by the administration of low doses of rapamycin, an inhibitor of mTORC1, resulted in an enlarged memory population, which more rapidly became central-memory in phenotype (71). Administration of rapamycin appears to be working by enhancing the Eomes/T-bet ratio (72), which favors memory differentiation (73). Skewing the Eomes/T-bet ratio toward Eomes would likely enhance TCM cell emergence because Tbx21−/− CD8+ T cells become CD62Lhigh more rapidly (74). Our in vivo studies using CD25−/− CD8+ T cells and IL-15−/− mice affirm previous in vitro studies examining the ability of IL-2 and IL-15 to support the differentiation of TEM- and TCM-like CD8+ T cell populations, respectively (51, 52). In addition, a recent report similarly found that CD25low effector CD8+ T cells preferentially became TCM cells (62). IL-2 could be working through the Blimp1/Bcl6 axis, as Prdm1−/− CD8+ T cells acquire a CD62Lhigh phenotype more rapidly (75), whereas Bcl6−/− CD8+ T cells have a decreased frequency of TCM cells (76). Furthermore, high levels of IL-2 are known to enhance Blimp1 expression while repressing Bcl6 expression in vitro (64). CD4+ T cell help is also known to result in decreased CD62L expression on Ag-specific CD8+ T cells (77). We hypothesize that this effect is likely due to IL-2 production by the “helping” CD4+ T cells (57, 78), which will then modulate CD62L expression. Our data demonstrated a heretofore unappreciated linkage between Ag availability and IL-2/IL-15 cytokines in the regulation of CD62L expression.
In summary, our results demonstrated that the differentiation of effector and memory CD8+ T cell populations occurred very early in the immune response to pathogens. At these early time points, signals generated by TCR engagement, costimulatory molecules, and the cytokine milieu are integrated by the responding CD8+ T cells to shape the development of effector and memory subsets. Thus, a thorough understanding of the early inflammatory environment and the cellular sources generating this environment will play a critical role in understanding the development of immunological memory.
We thank Dr. Charles Surh (The Scripps Institute) for providing CD25-deficient bone marrow. We thank Drs. Evan Jellison and Brian Sheridan for critically reading the manuscript and Kristina Gumpenberger and Quynh-Mai Pham for expert technical assistance.
This work was supported by National Institutes of Health Grants AI41576, AI76457, P01 AI56172 (to L.L.), and F32AI074277 (to J.J.O.).
Disclosures: The authors have no financial conflicts of interest.