Viral infections have been shown to induce lymphopenias that lower memory CD8 T cell frequencies, and they also have been shown to cause a permanent loss of memory cells specific to previously encountered pathogens. In this study, the patterns and significance of virus-induced memory CD8 T cell depletion were examined in mice immune to heterologous (Pichinde, vesicular stomatitis, vaccinia) viruses and subsequently challenged with acute or persistent lymphocytic choriomeningitis virus infections. Memory CD8 T cell loss was comprehensive and occurred in both lymphoid and peripheral tissues of the immune host. The impact of the loss of memory T cells was reflected by in vivo cytotoxicity assays, which showed decreased clearance of epitope-expressing targets. Memory CD8 T cell loss occurred very early (day 2) after infection, and was thereafter sustained, consistent more with an active deletion model than with a competition model. Cross-reactive T cells, in contrast, increased in number, but memory cells were reduced whether or not there was competition from cross-reactive T cells. Memory T cell loss was more profound during persistent infection than after acute infection. Adoptive transfer studies showed that, unlike the resolved acute infection, in which the reduced memory frequencies became stable, memory T cell loss was a continuously ongoing process during persistent infection. This study therefore links an early virus-induced lymphopenia to a subsequent long-term loss of CD8 T cell memory and offers a new mechanism for immune deficiency during persistent viral infections.
Improved techniques for identifying virus-specific T cells have shown that CD8 T cell frequencies specific for a given epitope are often as high as 2–5% in the resting memory state and may remain stable for the lifetime of a mouse (1, 2). Although most reported frequencies are somewhat lower in humans, some EBV, CMV, HIV, and human T cell leukemia virus (HTLV)3 epitope-specific T cells are within this range during acute and persistent infections (3, 4, 5, 6). Because a host is exposed to many viral infections and other antigenic challenges during its lifetime, it should not be possible to maintain such high frequencies for every epitope, and it has been proposed that the host accommodates this pathogen-induced diversification of its memory pool not by increasing the size of its lymphoid organs, but instead by deleting pre-existing memory cells (1, 7). This deletion only occurs if the memory cells do not cross-react with the recently acquired pathogen; cross-reactive cells are retained or enriched in frequency (8). These conclusions are strongly supported by studies in mice documenting the loss of virus-specific CD8 memory cells after one or a series of subsequent infections with unrelated viruses (1, 7, 9), and similar conclusions have been made in a bacterial system (10). For example, lymphocytic choriomeningitis virus (LCMV)-specific memory cells are progressively lost in LCMV-immune mice after infections with vaccinia virus (VV), Pichinde virus (PV), murine CMV, and vesicular stomatitis virus (VSV) (1).
What is not known is when this depletion takes place, how widespread it is throughout the body, and whether it is manifested by a deletion in protective immunity. Each of these issues is important to address. The deletion may occur passively, when newly formed memory cells compete with previously formed memory cells in reconstituting the memory pool after termination of infection, or actively, if the pre-existing memory cells are directly killed off by some mechanism. In fact, there is evidence of considerable CD8 memory T cell apoptosis occurring in the wake of the virus-induced lymphopenia that is common during the early stages of many viral infections (11, 12). Type 1 IFN has been implicated in driving this apoptosis and loss in CD44highCD8+ T cells, but the relevance of this to the long-term maintenance of memory is uncertain (11).
New evidence has indicated that T cells in the periphery may be very different from those in lymphoid organs in regard to proliferative and apoptotic properties (13, 14). Indeed, they may resist apoptotic forces that silence the immune response in lymphoid organs (14), thereby allowing them to remain at high levels in the periphery following infection (13). Studies on the loss of memory T cells after infection have focused on the loss in the lymphoid organs and peripheral blood (1, 7, 9). Given the changes in migration and adherence properties of memory T cells that can be initiated by cytokines, it must be clarified whether there are reservoirs of memory T cells undetected by the methods used. A loss of functional CTL memory in vivo in terms of impaired clearance of Ag-expressing cells would argue that the loss in memory cells is comprehensive and occurs throughout the body. In this study, we will use in vivo cytotoxicity assays to demonstrate the loss of functional memory in both lymphoid and peripheral organs.
Most studies on the loss of memory have focused on the consequences of acute sterilizing infections that drive a massive expansion and then deletion of CD8 T cells, as the immune system responds to the pathogen and then contracts when the pathogen is cleared (1, 7, 8). What is less clear is what happens under conditions of persistent infection, when the immune system does not return to homeostasis because the persisting pathogen continues to pose an antigenic challenge (9). Persistent infections in humans with HIV, HTLV, hepatitis C virus (HCV), hepatitis B virus, CMV, EBV, and rubella virus chronically stimulate the immune system and have been associated with immune suppression (15, 16, 17, 18, 19). It is difficult to successfully vaccinate these immune suppressed individuals, but what happens to previously acquired memory T cells specific to other pathogens when the host subsequently acquires a persistent infection?
To address these issues, we have examined the impact of acute sterilizing infections with LCMV, strain Armstrong, or of persistent infections with its highly disseminating clone 13 variant, on the survival of memory T cells and functional memory to PV, VSV, and VV. We report in this study the severe and global attrition in memory T cell number and function during and after resolution of an acute infection and throughout a persistent infection. We also show that this attrition most dramatically affects T cells that do not cross-react between the viruses.
Materials and Methods
LCMV, strain Armstrong, and its highly disseminating variant, clone 13, are ambisense RNA viruses in the Old World arenavirus family and were propagated in BHK21 baby hamster kidney cells (20, 21). VSV, strain Indiana, a negative strand RNA virus, was propagated in BHK21 cells. VV, strain WR, a DNA virus in the pox virus family, was propagated in NCTC 929 cells (20). PV, strain AN3739, a New World arenavirus only distantly related to LCMV, was propagated in BHK21 cells (21).
Generation of virus-immune mice
Male C57BL/6 (B6) and B6.PL Thy-1a/Cy (Thy-1.1+) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) at 5–6 wk of age and maintained under specific pathogen-free conditions within the Department of Animal Medicine at the University of Massachusetts Medical School. To generate virus-immune mice, 6- to 8-wk-old mice were inoculated i.p. with 2 × 107 PFU PV or 5 × 105 VV or i.v. with 1 × 106 PFU VSV. Virus-infected mice were considered virus immune 6 wk or longer of infection. To induce acute or persistent LCMV infections in these virus-immune mice, a moderate dose (5 × 104 PFU) of LCMV-Armstrong was inoculated i.p., or a high dose (5 × 106 PFU) of LCMV-clone 13 was inoculated i.v. The high dose of LCMV-clone 13 was enough to establish a chronic virus infection, which we confirmed by the presence of virus in the serum (average 5 × 104 PFU/ml) 6 wk after infection.
In vivo cytotoxicity assay
In vivo cytotoxicity was determined using B6 splenocytes coated with peptide epitopes and differentially labeled with the fluorescent dye CFSE (Molecular Probes, Eugene, OR) (22). The target cells were prepared from naive B6 splenocytes and then divided into two or three groups with equal number of cells. Although the reference population was untreated, the target populations were incubated with the indicated peptide (1 μM, 0.5 μg/ml; 45 min at 37°C, 5% CO2). Reference cells and peptide-pulsed target cells were then labeled with different concentrations of CFSE (2 and 0.5 μM for two targets, or 2, 2/3, and 2/9 μM for three targets), combined, and injected i.v. into mice. Lymphocytes from spleens or lungs were isolated 24 h later, and the ratio of CFSElow/CFSEhigh cells was determined by flow cytometry. The percentage of specific killing was calculated as follows: (1 − (ratio immune/ratio naive)) × 100. Ratio = percentage of target/percentage of reference.
Isolation of lung lymphocytes
Lung lymphocytes were isolated using an adaptation of a previously described protocol (23). The lung vascular bed was first flushed with 10 ml of chilled HBSS (Life Technologies, Grand Island, NY) introduced via cannulation of the right ventricle of the heart. The excised lungs, separated from all the associated lymph nodes, were minced and incubated for 1 h at 37°C in 5 ml of HBSS (10% FBS) containing 125 U/ml collagenase I (Life Technologies), 60 U/ml DNase I, and 60 U/ml hyaluronidase (Sigma-Aldrich, St. Louis, MO). The resulting single cell suspension was layered over Lympholyte-M (Cedarlane Laboratories, Hornby, Ontario, Canada) and centrifuged at 500 × g for 20 min at 4°C. Then the lymphocytes were collected.
Intracellular IFN-γ staining
Single cell lymphocyte suspensions were prepared from spleens, lungs, and the peritoneal cavity. The erythrocytes were lysed using a 0.84% NH4Cl solution. PV or VSV peptide-specific, IFN-γ-secreting CD8+ T cells were detected using the Cytofix/Cytoperm Kit Plus (with GolgiPlug; BD PharMingen, San Diego, CA), as described previously (24). Briefly, cells were incubated with 5 μM of synthetic peptide, 10 U/ml human rIL-2 (BD PharMingen), and 0.2 μl of GolgiPlug for 5 h at 37°C. For VV-specific CD8 T cells, instead of peptides, splenocytes were incubated with VV-infected peritoneal exudate cells (PECs) (10:1 ratio) isolated from day 2 VV-infected mice. To visualize overall functional memory CD8 T cells, cell were incubated with 1 μg of purified anti-mouse CD3ε mAb (145-2c11; BD PharMingen). Following preincubation with 1 μl of Fc block (2.4G2) in 96-well plates containing 100 μl of FACS buffer (HBBS, 2% FCS, 0.1% NaN3), the cells were stained for 30 min at 4°C with combinations of fluorescently labeled mAbs specific for CD44 (IM7, FITC), CD8α (53-6.7 PerCP), Thy-1.1 (OX-7, PE), and IFN-γ (XMG1.2, allophycocyanin), all purchased from BD PharMingen. Freshly stained samples were analyzed using a BD Biosciences FACSCalibur and CellQuest Software (San Diego, CA). In flow cytometry plots in Figs. 1, 3, and 6, each row represents the profile of a single mouse.
Several previously defined T cell epitopes encoded by the various viruses were used in this study. These included the immunodominant PV epitopes nuclear protein (NP) (38–45) (SALDFHKV), NP (122–132) (VYEGNLTNTQL), and one weak subdominant epitope NP (205–212) (YTVKFPNM) (8). The NP (122–132) is restricted by H-2Db molecule, and NP (38–45) and NP (205–212) are restricted by H-2Kb. The subdominant PV NP (205–212) epitope shares 6 of 8 aa with the subdominant LCMV NP (205–212) (YTVKYPNL) epitope, differing only in the Kb-binding motifs. The close resemblance between two epitopes from PV and LCMV leads to CD8 T cell cross-reactivity between the two viruses (8). For instance, when PV- or LCMV-immune mice are challenged with the heterologous virus, subdominant NP205-specific CD8 T cells are preferentially expanded and become dominant over otherwise dominant epitope-specific CD8 T cells (8). Other LCMV epitopes used included the Db-restricted gp33–41 (KAVYNFATC) and gp276–286 (SGVENPGGYCL) epitopes (25, 26). For visualization of VSV-specific CD8 T cells, the Kb-restricted VSV N protein-derived epitope (52–59) (RGYVYQGL) was used (13). Synthetic peptides listed in this work were purchased from American Peptide (Sunnyvale, CA) and were purified with reverse-phase HPLC to 90% purity.
Peptide-loaded H-2Kb and H-2Db tetramers were prepared, as previously described (27). For staining, cells were first blocked against nonspecific binding with an Fc block (2.4G2; BD PharMingen) and unlabeled streptavidin (Molecular Probes). Samples were washed and then costained with the indicated peptide-loaded tetramer and fluorescently labeled mAbs specific for CD44 and CD8α (53-6.7 PerCP) for 1 h at 4°C.
Adoptive transfer of PV-immune splenocytes
Spleen cells were prepared from PV-immune B6.PL Thy-1a/Cy (Thy-1.1) mice and labeled with 2 μM CFSE (Molecular Probes) solution for 15 min at 37°C. A total of 2 × 107 PV-immune cells was then injected i.v. into Thy-1.2 hosts. Three weeks after transfer, spleens were removed from host mice, and donor CD8 T cells were visualized using Thy-1.1 (OX-7, PE) mAb (28).
Attrition of PV-specific CD8 T cells upon acute or persistent LCMV infection
We first designed experiments to follow the fate of noncross-reactive and cross-reactive PV-specific memory CD8 T cells on heterologous acute or persistent LCMV infection. We examined two immunodominant PV-encoded epitopes (NP38, NP122) that do not significantly cross-react with LCMV, and a third subdominant epitope (PV-NP205) that shares 6 of 8 aa and cross-reacts with a subdominant LCMV epitope (LCMV-NP205) (8). Fig. 1,A shows, by peptide-induced intracellular IFN-γ assays, that CD44highCD8+ T cells from C57BL/6 mice previously infected with and now immune to PV exhibited a characteristic immunodominance profile in the memory pool, with NP38 > NP122 > NP205. PV-immune mice were then subsequently infected with LCMV-Armstrong or LCMV-clone 13 to establish acute or persistent infections, respectively. Six weeks after the LCMV infections, the mice were then examined for the status of both noncross-reactive and cross-reactive PV-specific CD8 T cells (Fig. 1,A). Substantial attrition of noncross-reactive NP38- and NP122-specific CD8 T cells was evident in the spleens after resolution of the acute LCMV infection and during the persistent LCMV infection. In acute LCMV-infected mice, compared with PV-immune mice, only ∼30% of NP38- and NP122-specific CD8 T cells survived the infection. The level of attrition, however, was far more profound in LCMV-clone 13-infected mice, which had lost >90% of those PV-specific T cells. This attrition of noncross-reactive memory CD8 T cells was also evident when calculated into absolute numbers per spleen and was highly significant (p < 0.05) (Fig. 1,B). Although PV-immune mice and PV + LCMV-Armstrong-immune mice had comparable spleen sizes and cellularity, clone 13-infected mice had smaller spleens, and this overall lymphopenia further contributed to the severe attrition (Fig. 1 B).
Recoveries of the cross-reactive NP205-specific CD8 T cells were quite different from the noncross-reactive ones. Our previous reports demonstrated that an acute LCMV-Armstrong infection preferentially expanded cross-reactive NP205-specific T cells from PV-immune mice, rendering them immunodominant over other LCMV epitope-specific CD8 T cells (8). Consistent with those results, a significant 22-fold increase in the NP205-specific CD8 T cells was observed in PV-immune mice after resolution of an acute LCMV-Armstrong infection (Fig. 1 A). We questioned what would happen after infection of PV-immune mice with LCMV-clone 13, which encodes the cross-reactive epitope, but which normally deletes or anergizes T cells responding to its own epitopes (29). Two critical questions can be addressed in this specific sequence of virus infection. First, would the recall cross-reactive NP205 response provide protective immunity against LCMV-clone 13 infection and prevent the host from acquiring the persistent LCMV infection? Second, would memory CD8 T cells be as easy to clonally exhaust as naive T cells responding to this persistent infection? To address these questions, we first performed plaque assays with serum from each group of PV-immune mice 6 wk after LCMV-clone 13 infection. High virus titers were detected (5–6 × 104 PFU/ml serum), indicating that high dose clone 13 infection was able to establish persistent virus infection in PV-immune mice. There was a modest increase in NP205-specific CD8 T cells in the spleen of clone 13-infected mice, but the frequencies after infection with this clonally exhausting strain of virus were substantially (∼7-fold) lower than those arising after the acute LCMV-Armstrong infection.
CD8 T cell attrition measured by peptide-tetramer complexes
One limitation inherent to IFN-γ assays is that only functionally competent memory CD8 T cells are scored as positive. Therefore, it was possible that the functionality of some of the PV-specific memory CD8 T cells may have been compromised by the LCMV infections, especially under conditions of persistent infections expressing a cross-reactive epitope. We, thus, in another experiment, quantified the cross-reactive (NP205) and noncross-reactive (NP38) PV-specific CD8 T cells with MHC tetramers (Fig. 1,C) and found patterns of attrition similar to those seen with the IFN-γ assay. In this experiment, LCMV-specific CD8 T cells were also visualized by tetramer. Previous reports have shown that even during the clonal exhaustion induced by persistent LCMV infection, some LCMV-specific CD8 T cells could still be detected via tetramer, although their effector functions appeared to be compromised (29, 30). We reasoned that severe attrition of PV-specific CD8 T cells in clone 13-infected mice might be partially due to a dilution effect by excessive accumulation of the anergized, nonfunctional LCMV-specific CD8 T cells. However, the data did not support that notion. In comparison with PV + LCMV-Armstrong-immune mice, comparable levels of gp276-specific CD8 T cells and significantly fewer gp34-specific CD8 T cells were detected after LCMV-clone 13 infection (Fig. 1 C). The data are consistent with the notion that severe attrition of memory CD8 T cells by persistent virus infection is not due to simple dilution or to anergization, but instead to a physical deletion of pre-existing memory CD8 T cells.
Memory T cell attrition in peripheral tissues
Viral infections may cause a redistribution of memory T cells and migration of cells from lymphoid organs to peripheral tissues (31), and it has been shown that, during the persistent LCMV infection, while the majority of LCMV-specific CD8 T cells are deleted from the spleen, significant numbers of LCMV-specific CD8 T cells remain viable in the peripheral organs (30). We therefore examined the degree of PV-specific memory CD8 T cell attrition in the peritoneal cavity (Fig. 2,A) and lung (Fig. 2 B) after acute or persistent LCMV infection. The results revealed similar (PEC) or even more dramatic (lung) levels of attrition of noncross-reactive PV-specific CD8 T cells after the heterologous LCMV infections. There was even more severe attrition in the peripheral tissues after the LCMV-clone 13 persistent infection than after the acute LCMV-Armstrong infection.
Preferential expansion of the cross-reactive NP205-specific CD8 T cells was also seen in the lung and the PEC after LCMV-Armstrong infection, closely resembling that of the spleen. However, unlike the spleen of LCMV-clone 13-infected mice, the clonal exhaustion of NP205 CD8 T cells appeared to be incomplete in the PEC, as significant numbers of NP205-specific CD8 T cells still survived in the peripheral tissues (Fig. 2 B).
Attrition of VSV- and VV-specific CD8 T cells on acute or persistent LCMV infections
Our examinations of the attrition of PV-specific memory CD8 T cells after LCMV infections were complicated by the fact that the strongly cross-reactive NP205-specific CD8 T cell responses may have influenced the attrition pattern of noncross-reactive memory CD8 T cells. We therefore examined the attrition phenomenon in other viral infection sequences, in which any such cross-reactivity was less dramatic. VSV is a small RNA virus from which only one H-2b-restricted CD8 T cell epitope has been identified (VSV N protein 52–59) (13), and there is no known cross-reactivity between this sequence and the known LCMV epitopes. VSV-immune mice were therefore infected with either LCMV-Armstrong or LCMV-clone 13, and 8 wk later VSV NP-specific CD8 T cells were quantified via an IFN-γ assay (Fig. 3 A). As with PV-immune cells, there was significant attrition of VSV-specific CD8 T cells in both LCMV-Armstrong (∼3-fold)- and LCMV-clone 13 (∼10-fold)-infected mice. Again, the persistent infection resulted in a much more profound attrition.
In this study, we also used an anti-CD3-stimulated IFN-γ assay to quantify the total frequencies of memory CD8 T cells, regardless of specificity (Fig. 3 A). Compared with VSV-immune mice, significantly higher (33 vs 6.6%) frequencies of IFN-γ-producing cells were detected in VSV+LCMV-Armstrong mice upon anti-CD3 stimulation, probably due to the presence of many LCMV-specific memory CD8 T cells now contributing to the total memory cell number; higher proportions of the CD8 T cells after acute LCMV-Armstrong infection also costained with the CD44 memory phenotype marker. In contrast, in LCMV-clone 13-infected mice, the percentage of IFN-γ-producing cells upon anti-CD3 stimulation was far less than the LCMV-Armstrong -immune host and more comparable to that in VSV-immune mice. This is probably because the majority of LCMV-specific CD8 T cells have been physically deleted or anergized during the persistent LCMV infection (29). These data indicate that this severe attrition of VSV-specific CD8 T cells may occur without a major increase in number of newly generated functional CD8 T cells. This supports our view that attrition is not simply due to competition, but may involve an active deletion process.
VV is a large DNA pox virus used to vaccinate against small pox virus. With so many humans now persistently infected with HIV and other viruses, we questioned how immunity to VV, a virus thought incapable of causing even a low grade persistence, would fare in the wake of acute or persistent infections. It may be that this attrition process is partly responsible for the loss of vaccine efficacy several years after vaccination. To address the issue, VV-immune mice were infected with acute or persistent LCMV. We have previously shown that while LCMV-immune mice are partially protected from VV infection, VV-immune mice were not protected from LCMV infection (32). Any cross-reactivity between VV and LCMV was insufficient for LCMV infection to expand detectible populations of VV-specific memory populations (28). VV-specific CD8 T cells were visualized and quantified by incubating spleen cells from VV-immune mice with PEC from day 2 VV-infected mice (Fig. 3 B). These PEC display viral Ags and stimulate T cells in intracellular IFN-γ assays in a similar manner to VV-infected fibroblasts (33). Similar patterns of VV-specific CD8 T cell attrition were observed as with PV or VSV, and again, the level of attrition was far more profound upon LCMV-clone 13 persistent infection.
Loss of functional memory in vivo
It was once thought that memory T cells were functionally dormant and needed time for reactivation to exert protective immunity. However, accumulating data have revealed that some freshly isolated memory CD8 T cells can kill target cells within 4 h in vitro (34, 35), and the newly developed in vivo cytotoxicity assays have shown some lytic activity within 4 h in vivo (36, 37). Immediate killing activity, provided by memory CD8 T cells, is thought to be a factor for efficient host protection upon re-exposure to the same pathogen, and reduced memory CD8 T cell frequencies may compromise the protective immunity of an immune host (10, 38). We thus examined the functional consequences of memory CD8 T cell attrition by using in vivo cytotoxicity assays. Two target populations, labeled with different concentrations of CFSE, were adoptively transferred into naive, VSV-immune, VSV + LCMV-Armstrong-immune, or VSV + LCMV-clone 13 persistently infected mice. Host peripheral blood cells were isolated 24 h after transfer, and relative killing was measured by comparing the VSV N protein peptide-labeled population with the reference population (Fig. 4). Substantial (90%) killing activity against VSV N protein peptide-coated population was seen in VSV-immune mice compared with a naive host, but this in vivo killing was significantly compromised after either acute (65%) or persistent (<20%) LCMV infections. The more profound loss in in vivo killing observed after LCMV-clone 13 persistent infection reflected the reduced frequency of VSV-specific T cells present in those mice (Fig. 3 A).
Kinetics of memory T cell attrition after acute and persistent infections
All of the experiments described to date examined virus-specific memory T cell attrition 6–8 wk after a heterologous viral infection. In an attempt to clarify why and when attrition took place, we examined the kinetics of noncross-reactive (NP38) and cross-reactive (NP205) PV-specific CD8 T cell frequencies via peptide-induced intracellular IFN-γ assay upon acute or persistent LCMV infection by analyzing the peripheral blood of the infected mice at multiple time points, beginning on day 6 (Fig. 5 A). Both infections initially caused a dramatic decline in the noncross-reactive NP38-specific CD8 T cells, from 7% to less than 1%, by day 9. At this point, the level of attrition was comparable between LCMV-Armstrong- and clone 13-infected mice. The acute Armstrong infection allowed for a gradual partial recovery of the percentage of NP38-specific T cells, as the LCMV-specific CD8 T cell response subsided. However, there was much less recovery in the clone 13-infected host, perhaps because of the continuous immune response reflecting the continued presence of Ag. Thus, part of the reason for the more dramatic attrition of memory during persistent infection may be the failure of the host to allow for a homeostatic recovery phase during the persistent infection.
The kinetics of the cross-reactive NP205 response was quite different (Fig. 5 A). There was a substantial expansion of NP205-specific CD8 T cells after the LCMV-Armstrong infection, which resulted in high frequencies of these cross-reactive T cells in the memory state. The persistent LCMV-clone 13 infection also caused an initial expansion of these T cells, but this was soon followed by down-regulation, perhaps as a consequence of apoptosis and anergization of the LCMV-specific T cells that occur in this system. It is noteworthy that the modest frequencies of NP205-positive T cells remaining after clone 13 persistent infection reflect a very substantial attrition of those cells after their transient peak earlier in infection.
We questioned how these LCMV infections altered the ability of T cells in vivo to lyse NP38- or NP205-coated targets. NP38-coated targets were efficiently lysed in vivo when inoculated into PV-immune mice (Fig. 5 B). Interestingly, however, there was little lysis of NP205-coated targets, perhaps reflecting the relatively low frequencies of NP205-specific T cells in PV-immune mice. This pattern changed dramatically in PV-immune mice after resolution of an acute LCMV infection, which deletes the NP38-specific T cells and expands the cross-reactive NP205-specific T cells in the memory pool. Now, the NP205-coated targets were lysed in vivo, and the lysis of the NP38-coated targets was much reduced. After clone 13 persistent infection, there was poor lysis against either of the targets.
Very early (day 2) loss of CD8 memory
The data shown in Fig. 5 A revealed that significant attrition of memory T cells may occur at stages of infection before there is a substantial increase in T cells specific to the infecting virus. This led us to consider the potential role of virus-induced lymphopenia in a lasting attrition of memory T cells. Virus-induced lymphopenia is a common feature of severe human and animal viral infections, including measles (39), influenza (40), Ebola (41), Venezuelan equine encephalitis (42), varicella zoster (43), and LCM viruses, among others (11). The lymphopenia occurs early during LCMV infection when viral titers are high and is associated with a type I IFN-dependent bystander apoptosis and loss of mature peripheral CD8 T cells, particularly within the memory type CD44high subset (11). This lymphopenia does not require infection of the T cells by virus, and, in fact, LCMV does not infect CD8 T cells.
We therefore questioned what role this bystander apoptosis occurring during the virus-induced lymphopenia plays in the long-term attrition of memory CD8 T cells upon heterologous virus infections. PV-immune mice were infected with either LCMV-Armstrong or LCMV-clone 13 and examined for memory T cell loss 2 days after infection, when virus-induced lymphopenia is apparent. PV-specific NP38 or NP122 CD8 T cells were then visualized and quantified via IFN-γ assay in the context of the memory marker, CD44. Fig. 6,A shows that the virus-induced lymphopenia appeared to preferentially reduce memory phenotype CD44high CD8 T cells, because 44% of the CD8 T cells were CD44high in the PV immune, but this was reduced to 32 and 34% after the LCMV infections. Thus, the ratio between CD44low and CD44high subsets changes as a consequence of the lymphopenia, as we noted previously (11, 12) (Fig. 6,A). Reductions of the noncross-reactive PV-specific memory CD8 T cells were seen at day 2 after either viral infection (Fig. 6 A). The decrease in PV-specific CD8 T cells was also even more apparent when the percentage was translated into absolute numbers per spleen, as there was overall reduction in the size of the spleens of the virus-infected mice at this point (NP38: PV-immune, 5.8 ± 2.5 × 105; PV + LCMV-Armstrong, 2.4 ± 1.3 × 105; PV + LCMV-clone 13, 1.5 ± 0.4 × 105). We assume that these observed reductions in virus-specific T cells as monitored by intracellular IFN-γ production indicate a loss in T cell number instead of just function, because there is a loss in the total number of CD44highCD8+ cells, and our previous work showed reductions in MHC tetramer-defined memory cells when poly(I:C) was used as a surrogate for virus-induced cytokine stimulation (11). However, it also remained possible that this disappearance of virus-specific CD8 T cells from the host spleen at day 2 may not be due to their apoptosis, but could be due to their migration into other sites or, as a consequence of structural changes in the spleen, they may be entangled with the parenchymal tissues in ways that made the lymphocytes difficult to isolate by our conventional methods. We think this unlikely, as we were unable to detect an accumulation of memory cells in other sites or in tissues such as liver, lung, or fat, in which proteolytic enzymes are used as part of the extraction process (data not shown).
Nevertheless, we sought to clarify the extent of the total body deletion in functional memory cells at 2 days postinfection by using the in vivo cytotoxicity assay. We reasoned that if Ag-specific CD8 T cells still existed somewhere within the host with functions intact, they should be able to kill implanted Ag-expressing cells. PV NP38 peptide-labeled target cells along with the reference target cells were transferred into naive, PV-immune, and PV-immune mice infected with either LCMV-Armstrong or LCMV-clone 13 2 days previously. Fig. 6,B shows that the NP38-coated targets were efficiently eliminated in the spleen of PV-immune mice, but this elimination was greatly compromised in the PV-immune mice infected for 2 days with either strain of LCMV. We could not find reservoirs in the body in which the LCMV infections did not reduce the PV-specific clearance of cells. Fig. 6 C shows the survival of implanted cells in the lung, and these results were similar to that of the spleen. The data indicate that CD8 T cell attrition and its functional consequences occur very early after infection and occur comprehensively throughout the host.
Attrition of PV-specific memory CD8 T cells at later stages of persistent infection
The experiments above documented a significant attrition of noncross-reactive memory CD8 T cells early after viral infection and lasting for months thereafter, but we questioned whether this attrition occurred only during the early stages of infection or would be continuous during a persistent infection. To address this phenomenon, we used an adoptive transfer system, in which CFSE-labeled splenocytes from Thy-1.1+ PV-immune mice were transferred into Thy-1.2+ congenic hosts, and their numbers and division were monitored 3 wk later. Fig. 7,A (top) shows the total number of CD8 donor cells and (bottom) the donor CD8-gated CSFE-labeled NP38-specific T cells. In a control-uninfected C57BL/6 mouse host, ∼7% of the donor CD8 T cells were specific for NP38 and about one-half of them (3%) showed signs of cell division, i.e., loss of CSFE label. When cells were transferred into uninfected IL-15 knockout mice, there was a moderate reduction in NP38-specific T cell frequencies to 4% and no evidence of division. This suggests that the low levels of division of memory CD8 T cells in naive mice were likely to be an IL-15-dependent homeostatic division, as would be predicted from other work (44, 45). When PV-immune splenocytes were transferred into an LCMV-Armstrong-immune host, the fate of the donor cells was nearly identical with that in the uninfected control host. Virus gets cleared from the host by 8 days after acute LCMV-Armstrong infection, and thereafter, there is stable homeostasis, just as in an uninfected host. In contrast to the LCMV-immune mice, transfer of PV-specific memory T cells into a host harboring an LCMV-clone 13 persistent infection for 42 days led to a dramatic reduction in NP38-specific donor cells. This attrition in the persistently infected host was not due to the lack of homeostatic cell division, as many (72%) of the remaining PV-specific CD8 T cells lost CSFE label. This most likely indicates that any cell division is offset by greater levels of apoptosis, driving the significant attrition of memory cells in this persistently infected environment. Fig. 7 B shows the total numbers of donor NP38-specific memory cells in these different host environments and further documents the severe attrition in the persistently infected host. These data therefore indicate that chronic virus infections can induce a continuous attrition of noncross-reactive memory CD8 T cells long after the initial virus infection has occurred.
In this study, we have documented that both acute and persistent LCMV infections led to profound depletion of pre-existing noncross-reactive CD8 T cells specific to PV, VSV, and VV, and that this depletion resulted in significant functional consequences reflected in the decreased clearance of Ag-expressing target cells in vivo. This depletion begins very early in infection, in association with a previously described virus-induced type 1 IFN-dependent lymphopenia affecting CD44highCD8+ cells (11, 12). In the acute sterilizing infection, there was a modest, but incomplete rebound in memory cell frequency, after which they were stably maintained. During the persistent infection, there was very little rebound, and the attrition occurred continually, as shown by adoptive transfer studies (Fig. 7). This indicates that a persistent infection may cause a much greater deletion in memory T cells than an acute infection, perhaps because in the persistent infection the attrition is a long-term ongoing process.
The results also indicate that the CD8 memory T cell attrition is comprehensive, occurring throughout the body, as reflected in reduced frequencies in the spleen, PBL, PEC, and lung (Fig. 2), and by the impaired in vivo cytotoxicity in the spleen, PBL, and lung (Figs. 4,B and 6 B). Hence, any reductions in memory T cells occurring as a consequence of subsequent infections cannot be attributed simply to their migration to different sites in the host.
The profound nature of this T cell attrition, along with the compromised ability to clear Ag-expressing target cells, argue that this is an important process that could compromise immunity in an individual. This leads one to question how persistent human infections such as HIV, HCV, HTLV, hepatitis B virus, CMV, EBV, and rubella may compromise immunity to previously encountered pathogens. This is an area that has not been carefully examined. It is not known whether human memory cells are reduced after infection to the degree that murine memory cells are. It is possible that the larger immune system of humans has a greater buffering capacity, allowing it to accommodate many more memory cells than does the mouse, and work along these lines needs to be clarified. It is known, however, that persistent human infections can cause immune suppression and compromise immunity to previously encountered Ags, but the mechanisms remain unclear (15).
One of the problems with studying the impact of human persistent infections on memory to previously encountered viruses is that many of those encountered human viruses persist at low levels or are continually re-encountered in the environment. Should an infection such as HIV cause a reduction in memory to a latent virus that has never left the host, the latent virus may reactivate and stimulate a secondary T cell response (19). This would complicate the interpretation of quantitative differences of T cells, a problem that is not a factor in the mouse studies.
Application of these findings to human systems is further complicated by the potential of different viruses to cross-react at the T cell level. In this study, we show that the rules that apply to noncross-reactive T cells do not apply to cross-reactive ones. Instead of an attrition, the cross-reactive T cells tend to be preserved or even enriched after infection with another virus, as shown with the NP205 epitopes that are cross-reactive between PV and LCMV. Work is just now showing substantial CD8 T cell cross-reactivity between divergent human viruses, such as influenza and HCV (46), influenza and EBV (47), corona and papilloma (48), etc., so these potentials for cross-reactivity would need to be clarified before a comprehension of this phenomenon in humans could be made.
We have formulated two possible models for CD8 T cell attrition during heterologous virus. The passive competition model proposes that newly generated virus-specific CD8 T cells dilute pre-existing memory CD8 T cells and then compete for structural niches and/or nutrients. The active deletion model proposes that pre-existing memory CD8 T cells are actively killed off by factors released during the virus infection. These two models are mechanistically distinct, but not necessarily mutually exclusive, but we can conjecture which model better fits our observations. In terms of the kinetics of T cell attrition, while the passive model predicts a slow and gradual loss of pre-existing memory T cells in the wake of the developing new T cell response over the course of heterologous virus infection, the active deletion model predicts a loss of memory T cells unrelated to the accumulation of new cells. The data indicated attrition occurs abruptly by 2 days after virus infection, before the new virus-specific T cells have developed (Figs. 1,c and 6). According to the passive competition model, the survival of pre-existing memory cells will be dependent on the frequency of newly generated memory cells competing with them. The depletion of pre-existing memory CD8 T cells takes place before the expansion of new virus-specific CD8 T cells, and this depletion occurs more dramatically in the persistently infected mice, which have a lower T cell number than the acutely infected mice. To date, our observations favor the active deletion model over the passive competition model, but we feel that more work is needed to completely clarify these phenomena.
Virus-induced lymphopenia is a common feature of many virus infections, including LCMV (39, 11). In Figs. 5 and 6, we demonstrated its association with the depletion of a significant portion of pre-existing functional PV-specific CD8 T cells. The long-term loss of memory CD8 T cells upon heterologous virus infections could be mainly caused by this lymphopenia during the early phase of virus infection. Memory phenotype (CD44high) CD8 T cells are more susceptible to apoptosis than naive CD8 T cells during the virus-induced lymphopenia, and consequently, their number preferentially decreases in the wake of heterologous virus infection (11, 12).
A series of studies have documented a bystander division of CD44high CD8 T cells in response to virus infection or poly(I:C) treatment, which induces cytokines that mimic the early phase of virus infection (49, 50). Why then can their number not be restored by homeostatic proliferation? This is probably the consequence of two phenomena. First, bona fide virus-specific memory CD8 T cells respond relatively poorly to homeostatic expansion signals created by lymphopenic space compared with other CD44highCD8+ T cell subsets (12). Second, the virus-induced lymphopenia is relatively short-lived and it is soon followed by massive expansion of Ag-driven new virus-specific CD8 T cells. Thus, the combination of competition from Ag-specific and from homeostatically responding cells may keep the memory T cells specific to previously encountered viruses from recovering to their former frequencies.
The persistent LCMV infection caused a much more severe attrition of pre-existing memory CD8 T cells than the acute infection, and a recent report has shown that a persistent mouse γ herpes virus infection reduced T cells specific to influenza (9). This may mean that memory T cell attrition is a profound threat to those harboring persistent infections that continually induce apoptotic cytokines and stimulate T cell responses. Our adoptive transfer studies showed that this attrition can be a continuous process occurring during chronic virus infection long after its initiation (Fig. 7). This could be caused by the continued production of IFN and other apoptotic cytokines during the persistent infection or to the fact that there are activated T cells in persistently infected mice expressing proapoptotic signals such as Fas ligand, which may sensitize T cells to a bystander apoptosis (51). This needs to be clarified.
Memory T cells provide better protection against secondary challenge because they are present at high frequency, they are poised to rapidly activate and exert cytotoxic and cytokine-producing functions, and they are located in many nonlymphoid tissues, including mucosal sites to which many pathogens gain access (13, 14, 31, 34). Those memory CD8 T cells in the peripheral tissues have been hypothesized to play a role as a first line of defense against infection. We show that attrition of memory CD8 T cells is comprehensive and occurs throughout the whole body, and that this reduced number of virus-specific CD8 T cells is associated with impaired cytotoxicity and clearance of APCs in vivo (Figs. 2 and 5). This depletion of functional memory CD8 T cells seems sufficiently profound to have significant impact on protective immunity, as predicted by other studies correlating CD8 T cell frequencies with protective immunity in viral (38) or tumor (10) systems.
We thank Keith Daniels, Michael Brehm, and Liisa Selin for reagents and support. We thank Jacques Peschon and Joosoo Kang for IL-15 KO mice.
This study was supported by National Institutes of Health Grants AR35506, AI46578, and AI057330-01. The contents of this publication are solely the authors’ and do not represent the official view of the National Institutes of Health.
Abbreviations used in this paper: HTLV, human T cell leukemia virus; HCV, hepatitis C virus; LCMV, lymphocytic choriomeningitis virus; NP, nuclear protein; PEC, peritoneal exudate cell; PV, Pichinde virus; VSV, vesicular stomatitis virus; VV, vaccinia virus.