Activated Antigen-Specific CD8⁺ T Cells Persist in the Lungs Following Recovery from Respiratory Virus Infections¹

Although humoral immunity provides complete protection against secondary challenge with homologous virus, it is ineffective against serologically distinct viruses (1, 2, 3). In contrast, cellular responses to cross-reactive epitopes (often from internal viral proteins) provide a substantial degree of protection against serologically distinct viruses (4, 5). For example, in the mouse model, protective cellular immunity can be demonstrated between influenza virus variants that differ in coat proteins, but share internal proteins (6). This form of immunity, sometimes referred to as heterosubtypic immunity, is not able to prevent re-infection per se, but can reduce the maximal viral load, mediate faster viral clearance, and provide a substantial degree of protection against challenge with a lethal dose of virus in animal models (5, 7, 8, 9, 10, 11, 12). Mouse studies indicate that heterosubtypic immunity is mediated by both CD4⁺ and CD8⁺ memory T cells, although the CD8⁺ subset is generally considered to be the more important (7, 13).

The recent development of MHC class I tetramers has made it possible to directly identify Ag-specific memory T cells ex vivo by flow cytometry (14, 15, 16). This approach has revealed that memory CD8⁺ T cells are maintained at unexpectedly high frequencies in the spleen (12, 16, 17). For example, in C57BL/6 mice that have recovered from a primary Sendai virus infection, memory T cells specific for an immunodominant nucleoprotein (NP)³ epitope (Sen-NP_324–332/K^b) can be detected at frequencies as high as 6% of splenic CD8⁺ T cells (17). This is much higher than the frequencies typically obtained by classical limiting dilution analysis, which are normally about 0.3% of splenic CD8⁺ T cells (18, 19, 20). Similar frequencies have been obtained in an influenza virus system using tetramers corresponding to the influenza immunodominant T cell epitope (Flu-NP_366–374/D^b). Flu-NP_366–374/D^b-specific memory T cells induced by intranasal influenza virus infection comprise about 1% of the total CD8⁺ T cells in the spleen (12, 17).

Recent studies have documented a number of crucial differences between naive and memory T cells subsets. Naive T cells most closely resemble mature thymocytes and are considered to be Ag-inexperienced (21, 22). Memory CD8⁺ T cells can be distinguished from naive T cells on the basis of several characteristics as follows: 1) they respond rapidly to recall Ags, 2) they produce a different array of cytokines, 3) they have low costimulatory requirements, 4) they have a relatively low susceptibility to apoptosis, 5) they express high levels of adhesion molecules, and 6) they express low levels of the lymph node homing receptor (CD62 ligand (CD62L)) (22, 23). In addition, memory CD8⁺ T cells are generally considered to be of a resting phenotype, although some memory cells may retain effector function (16, 24, 25, 26, 27, 28). Recent studies have also revealed that there is substantial heterogeneity among populations of memory cells with respect to cell turnover (29) and phenotype as measured by activation markers such as CD62L and CD45RA/B/C (21, 27, 29, 30). For example, memory CD8⁺ T cells in the spleen can be subdivided into two populations in terms of their level of CD62L expression, and there is a gradual switch from the CD62L^low to the CD62L^high phenotype over time (17, 23, 31).

Although memory CD8⁺ T cells play a central role in recall responses to infection, the presence of memory cells in the spleen and lymph nodes does not seem to correlate with cell-mediated protection. Elegant studies by Gerhard and colleagues (7) have shown that protective cellular immunity to influenza virus wanes rapidly and is substantially reduced within 3 mo of the primary infection. This is despite the fact that stable numbers of CD8⁺ memory cells persist in the secondary lymphoid system for over 1 year after infection (Ref. 31 and this paper). Similarly, protective CD8⁺ T cell responses to Sendai virus in B cell-deficient mice have been found to decline rapidly, despite substantial numbers of memory T cells in the spleen (D.L.W., unpublished data). Heterosubtypic immunity in humans is also generally considered to be weak, and this may reflect the waning of an initially strong response (5, 7, 32, 33, 34, 35, 36). The lack of a correlation between memory T cell numbers in the secondary lymphoid organs and protective immunity suggests that there may be a distinct pool of memory cells that mediates this function. Given previous reports for the persistence of intraparenchymal pulmonary lymphocytes (37, 38), we investigated whether memory cells could be detected in the lungs after viral clearance. The data show that Ag-specific T cells persisted in the lung tissue and airways for several months after the primary infection. Moreover, these cells persist in a highly activated state and are capable of proliferating and acquiring cytotoxic activity.

Sendai virus (Enders strain), influenza virus A/HK-x31 (x31, H3N2), and influenza virus A/PR8/34 (PR8, H1N1) were grown, stored, and titrated as previously described (18, 39). Female and male C57BL/6 mice were purchased from The Jackson Laboratory (Bar Harbor, ME) or the Animal Breeding Facility at the Trudeau Institute (Saranac Lake, NY) and housed under specific pathogen-free conditions. Mice (6–12 wk) were anesthetized by i.p. injection with 2,2,2-tribromoethanol and infected intranasally with 500 50% egg infectious doses (EID₅₀) of Sendai virus, 300 EID₅₀ of x31, or 3,000 EID₅₀ PR8 virus. Mice were considered to be “memory mice” when they had been infected with influenza or Sendai virus a minimum of 30 days previously. The OT-I mouse strain, on a C57BL/6 background (H-2^b) was originally obtained from Dr. Michael Bevan (University of Washington, Seattle, WA). These mice express a transgenic TCR specific for the SIINFEKL peptide of OVA (OVA_257–264) in the context of MHC class I, H2-K^b (40).

Sendai virus NP_324–332, influenza virus NP_366–374, and OVA_257–264 peptides were purchased from New England Peptide, Fitchburg, MA. Peptide purity was evaluated using reverse-phase HPLC analysis.

Single-cell suspensions were obtained from spleens and mediastinal lymph nodes (MLN) by passage through cell strainers, and spleen cells were additionally depleted of erythrocytes by treatment with buffered ammonium chloride solution. Bronchoalveloar lavage (BAL) cells were collected by lavage of the lungs three times with 1 ml HBSS. Cells derived from lung tissue were obtained by passing lavaged lungs through cell strainers. The cells were then resuspended in 80% isotonic Percoll and layered with 40% isotonic Percoll. After centrifugation at 400 × g for 25 min, the cells at the 80%/40% interface were collected, washed, and counted.

Cells were labeled with CFSE by incubation in HBSS containing 0.5–0.7 μM CSFE for 10 min in the dark. The cells were subsequently washed with HBSS or culture medium before use. Cultures were restimulated for 4 days in vitro with either the Sen-NP_324–332 or Flu-NP_366–374 peptides (0.1–0.5 μg/ml; or no peptide as an additional control) and 10 U/ml human rIL-2 (R&D Systems, Minneapolis, MN) at a cell density of 1 × 10⁶/ml in 24-well plates (17). In some experiments, peptide-pulsed, γ-irradiated (4000 rad) naive spleen cells were added as APCs.

MHC class I-peptide tetramers were generated by the Molecular Biology Core Facility at the Trudeau Institute as described previously (15). The two tetramers used in these studies (Sen-NP_324–332/K^b and Flu-NP_366–374/D^b) have been shown to be highly specific (Refs. 12, 17 , and 41 , and data not shown). Staining with tetrameric reagents was performed for 1 h at room temperature, followed by anti-CD8 tricolor (Caltag, Burlingame, CA) and either biotinylated or FITC-conjugated Abs specific for CD44, CD62L, CD69, CD25, or CD43 (BD PharMingen, San Diego, CA) on ice for 20 min. Stained samples were run on either a Becton Dickinson (San Jose, CA) FACScan or FACSCalibur flow cytometer and data was analyzed using CellQuest software (Becton Dickinson). In some experiments B cells were depleted before staining by panning on anti-Ig-coated flasks. The percentage of tetramer⁺ cells among total live cells was calculated by dividing the number of tetramer⁺/CD8⁺ events by the total number of events in a live cell gate. The absolute number of tetramer⁺ cells was calculated using this percentage and the number of cells isolated per mouse in each tissue as indicated by trypan blue staining. The accuracy of this method was confirmed by differential counting (data not shown). In some experiments, tetramer⁺/CD8⁺ T cells were isolated from the lung preparations by sorting on a Becton Dickinson FACStar^Plus flow cytometer.

Cytotoxicity assays were performed as described previously (20). Briefly, target cells (L-K^b or L-D^b transfectants) were labeled with ⁵¹Cr (Na₂CrO₄; New England Nuclear, Boston, MA) and then pulsed with peptide. Unpulsed target cells were used as negative controls. Various numbers of effector cells were incubated with 500 target cells for 5 h. The percentage of specific release was calculated using the following formula: percent specific release = (experimental − spontaneous)/(maximum − spontaneous) × 100. Spontaneous release values were obtained by incubation of target cells in complete tumor medium alone and were routinely <10% of maximum release. Maximum release values were obtained by the addition of 100 μl 1% Triton X-100.

Mice were killed by halothane inhalation, their tracheas were intubated, and their lungs were inflated with 1 ml of warmed Tissue Tech OCT embedding medium (Miles, Elkhart, IN). The trachea was then tied off, the lungs were frozen, and 5 μm frozen sections were cut, air-dried, and fixed in cold acetone for 5 min. Tissue sections were blocked with 2% normal rat serum and endogenous biotin and avidin was additionally blocked with the use of the Biotin/Avidin Blocking Kit (Vector Laboratories, Burlingame, CA). The sections were then incubated with biotinylated anti-CD8 (TIB210; BD PharMingen) followed by Alexa Fluor 488 (Molecular Probes, Eugene, OR). After washing the sections were then incubated with biotinolated anti-CD44 (BD PharMingen) followed by Alexa Fluor 594 (Molecular Probes). After rinsing, sections were coverslipped using Aqua Poly/mount mounting media (Molecular Probes), and viewed for fluorescence.

The relative frequencies of IFN-γ-secreting cells derived from spleen, lung, MLN, and BAL tissues were determined following stimulation with Sen-NP_324–332 or Flu-NP_366–374 in a standard ELISPOT assay system (42). Briefly, 96-well Multiscreen HA nitrocellulose plates (Millipore, Bedford, MA) were coated overnight at 4°C with 100 μl/well of rat anti-mouse IFN-γ (clone R4-6A2; BD PharMingen), at a concentration of 10 μg/ml. The plates were then washed and blocked before the addition of titered numbers of responding cells, irradiated (3000 rad) syngeneic normal spleen cells, peptide, and 10 U/ml human rIL-2. Plates were then incubated overnight at 37°C and developed overnight with a biotinylated detection Ab, rat anti-mouse IFN-γ (clone XMG1.2; BD PharMingen), followed by streptavidin-HRP (BD PharMingen) for 2 h at room temperature. Visible spots of IFN-γ-secreting cells were then enumerated using an Olympus (New Hyde Park, NY) SZH stereo zoom microscope system.

Effector T cells specific for the OVA_257–264/K^b-epitope were prepared by 4-day culture of naive CD8 cells from OT-1 mice as previously described (43). Briefly, naive CD8-enriched T cells were obtained by passing lymphoid cell suspensions through nylon wool columns and treating with anti-CD4 (RL172.4), anti-heat-stable-Ag (J11D), anti-MHC Class II (D3.137, M5114, CA4) mAbs and complement. Small resting CD8 T cells were harvested from Percoll gradients (Sigma, St. Louis, MO) and resuspended at appropriate cell concentrations in culture media. Naive CD8 cells were typically 90% pure as demonstrated by flow cytometry. APCs were enriched from spleens of normal C57BL/6 mice by anti-Thy1.2 (HO13.14 and F7D5), anti-CD4 (RL172.4), and anti-CD8 (3.155) mAbs and complement. T cell-depleted APCs were pulsed with 10 μM OVA_257–264 peptide for 30 min at 37°C and treated with mitomycin-C (Sigma) for an additional 30 min at 37°C. For Tc1 effector cell generation, naive CD8 T cells from OT-I-transgenic mice (2 × 10⁵ cells/ml) were stimulated with mitomycin C-treated OVA_257–264 peptide-pulsed APCs (6 × 10⁵ cells/ml) in the presence of 20 U/ml IL-2 (X63.IL-2 supernatants), 2 ng/ml IL-12, and 200 U/ml anti-IL-4 mAb (X63.Ag.IL4 supernatants). Alternatively, for Tc2 effector cell generation, naive CD8 T cells from OT-I-transgenic mice were stimulated with mitomycin C-treated OVA_257–264 peptide-pulsed APCs in the presence of 20 U/ml IL-2, 200 U/ml IL-4 (X63.IL-4 supernatants), and 20 g/ml anti-IFN-γ mAb (XMG1.2). Effector cell cultures were incubated for 4 days with additional 20 U/ml IL-2 added on day 2.

One × 10⁷ 4-day Tc1 or Tc2 effector cells were injected into syngeneic adult thymectomized, irradiated, bone marrow-restored mice as previously described (44). Such mice retain CD8 memory cells in almost constant numbers for many months. Mice in this study were used at 10 mo after adoptive transfer.

Previous studies have shown that intranasal Sendai virus infection of C57BL/6 mice induces memory T cells specific for the immunodominant Sen-NP_324–332/K^b epitope at high frequencies in the spleen and lymph nodes (ranging from 1 to 6% of CD8⁺ T cells) (17, 20, 45). These memory T cell populations are relatively stable and persist in these organs for over 1 year postinfection. To determine whether there was an additional long-lived reservoir of Sen-NP_324–332/K^b-specific T cells in the lungs, we analyzed the BAL of mice at various times after recovery from a Sendai virus infection. As shown in Fig. 1, high frequencies of Sen-NP_324–332/K^b-specific T cells could be readily detected in the lung airways (BAL) for many months after the virus had been cleared. For example, over 60% of the CD8⁺ T cells in the lung airways were specific for the Sen-NP_324–332/K^b epitope 44 days postinfection, which is 5 wk after viral clearance. This percentage declined slowly over the next year to around 16% at 1 year postinfection. In contrast, much lower, and more stable frequencies of Sen-NP_324–332/K^b-specific memory cells were found in the spleen (Fig. 1) and MLN (data not shown) throughout this time (ranging from 0.9 to 2.5%) (17). In terms of absolute cell numbers, there were generally about 1–4 × 10⁴ Sen-NP_324–332/K^b-specific T cells/mouse in the lung airways (BAL) about a month after the infection had been cleared (or ∼40 days postinfection). This number is remarkable, given that only 1–2 × 10⁵ Ag-specific T cells are normally present in the BAL at the peak of the acute CD8⁺ T cell response (day 10 postinfection), that virus is cleared from the system by day 8, and that inflammation in the lung resolves by 3 wk postinfection (as determined by histological analysis; A.G.H., unpublished data; Refs. 41 and 46). Although there is a steady decline in the numbers of Sen-NP_324–332/K^b-specific T cells in the airways (to <1,000/mouse at day 200 postinfection), substantial numbers of cells still persisted in the lung airways over 1 year postinfection (Fig. 2 A). In contrast, the absolute numbers of Sen-NP_324–332/K^b-specific T cells in the spleen are much higher (∼200,000/spleen) and does not vary significantly over the first year postinfection.

The persistence of Ag-specific T cells in the lung airways suggested that there was a pool of Ag-specific T cells in the lung tissue itself. To test this, leukocytes isolated from homogenized lung tissue (after lavage) were analyzed for the presence of Ag-specific T cells. As shown in Table I, remarkably large numbers of Sen-NP_324–332/K^b-specific T cells were identified in the lung tissue (∼61,000/mouse) on day 36 post Sendai virus infection (4 wk after virus clearance). The total number of Sen-NP_324–332/K^b-specific T cells in the lungs (tissue and airways combined) in this experiment was over 100,000/mouse, which is much greater than the numbers present in the local MLN (∼9,000), and almost half of the number of cells present in the spleen (∼218,000). In other experiments, we found that the absolute numbers of Sen-NP_324–332/K^b-specific T cells per mouse lung (tissue and airways) were variable, but always high, ranging from 30,000 to 120,000 per mouse. The absolute number of Sen-NP_324–332/K^b-specific T cells established in the lung (tissues and airways) at day 36 postinfection did not reflect the dose of virus used to infect the mice (data not shown).

A potential explanation for the presence of Ag-specific T cells in the lung tissue is that they are simply cells transiting through the lung in the circulation. However, several observations argue against this possibility. First, the absolute maximal number of tetramer⁺ cells in the total lung tissue that could be derived from the blood (based on mouse blood volumes, lymphocyte counts, and tetramer⁺ cell frequencies) would be about 100–300 (47). Second, the frequencies and phenotypes of cells isolated from the lung are distinct from those in the circulation. For example, frequency of Sen-NP_324–332/K^b-specific T cells in the blood is in the order of 0.1–0.3% of CD8⁺ T cells (D.L.W. and R.J.H., unpublished data), whereas the frequency of these cells in the lung ranges from 16 to 60% (Fig. 1). Third, immunohistochemical analysis indicated that most of the CD44⁺/CD8⁺ T cells in the lung were associated with the lung tissue, with very few in the capillaries (Fig. 3).

To assess the generality of the findings in the Sendai virus system, we next asked whether Ag-specific T cells persisted in the lungs following intranasal influenza virus infection. As shown in Table I, CD8⁺ T cells specific for the immunodominant Flu-NP_366–374/D^b epitope were readily detectable in the lung airways (BAL) and lung tissue long after virus had been cleared. The absolute number of Ag-specific T cells in the lung at day 45 (tissue and airways) was just under 10,000 and these cells represented 6.4% and 10.4% of CD8⁺ T cells in the lung tissues and airways respectively. These frequencies are similar to the frequencies of T cells specific for the immunodominant epitope in the acutely infected lung and the memory spleen (12, 17). The relatively low frequency of T cells specific for the Flu-NP_366–374/D^b epitope probably reflects the fact that multiple epitopes are involved in the response and that not all of them have been accounted for. In this regard, recent studies by Belz et al. (41) have identified a new CD8⁺ T cell epitope (Flu-PA_224–233/D^b) that is a component of the C57BL/6 T cell response to influenza virus infection. The total number of Ag-specific T cells in the lung declined with time and was <2,000 cells in the lung tissue and 200 cells in the BAL on day 90 postinfection (data not shown and Fig. 2,B). The slow decline in the numbers of Ag-specific T cells in the lung airways was similar to that observed in mice that had recovered from Sendai virus infection (Fig. 2, A and B). Secondary challenge of HKx31-primed mice with the serologically distinct influenza virus (A/PR8) greatly boosted the numbers of Ag-specific T cells present in all tissues. Thus, there were over 54,000 Flu-NP_366–374/D^b-specific T cells in the lungs at day 32 postsecondary infection (Table I). These data establish the generality of the results in the Sendai virus system.

Phenotypic analysis of Ag specific CD8⁺ T cells in the lungs of mice that had recovered from either a Sendai virus (40 days) or influenza virus (31 days) infection indicated that they were highly activated. For example, ∼70% of the Ag-specific T cells in the lung airways expressed CD69, which is a marker of recent activation (Tables II and III). In contrast, CD69 expression on Ag-specific memory cells in the spleen was substantially lower (20–30%). Similarly, the CD25 and CD43 activation markers were expressed by high frequencies of Ag-specific T cells in all tissues, although the levels were generally higher among Ag-specific T cells isolated from the lung airways (Tables II and III). CD43 has been reported to distinguish effector and memory cells and appears to be associated with cytolytic activity (48). However, in contrast to studies in the lymphocytic choriomeningitis virus system, CD43 was expressed on relatively high frequencies of Ag-specific T cells, long after viral clearance in both the lung and spleen, and, as discussed below, did not correlate with cytolytic activity. With increasing time, we observed that the frequencies of Ag-specific T cells expressing activation markers in the lung airways declined, although these numbers remained significantly higher than Ag-specific T cells in the spleen. For example, at 84 days post Sendai virus infection, the frequency of tetramer⁺ T cells in the lung airways that expressed CD69 had dropped from 72% (at day 40, Table II) to 35% (data not shown). In general, tetramer⁻/CD8⁺ T cells from all sites did not express a highly activated phenotype with the exception that tetramer⁻/CD8⁺ T cells in the lung airways were predominantly CD69⁺ (Tables II and III) (49, 50, 51).

Expression of the CD62L memory/activation marker was substantially different between Ag-specific T cells in the lung airways and spleen. Thus, Ag-specific T cells in the spleen were a mix of CD62L^low and CD62L^high cells and switched from being predominantly CD62L^low to being predominantly CD62L^high over time (Fig. 4), as previously described (31). In contrast, virtually all of the Ag-specific T cells in the lung airways were CD62L^low at all timepoints analyzed (Fig. 4). The frequency of CD62L expression by Ag-specific T cells in the lung tissue was always intermediate between the lung airways and spleen.

The general activation level of Ag-specific T cells in the lung is remarkable given the time that has elapsed since infectious virus was cleared and suggested that there may be a source of persistent Ag in the lung. To test this idea, we took advantage of a very sensitive LacZ T cell hybridoma assay for the detection of processed Sen-NP_324–332/K^b Ag expression ex vivo (52). However, processed Ag could not be detected in the lung at day 25 postinfection (data not shown) suggesting that other factors may be involved in the activation of the cells in vivo. This is consistent with other studies documenting the lack of persisting viral RNA in the lung 14 days after influenza virus infection (53).

We next asked whether Ag-specific T cells in the lungs expressed constitutive cytolytic activity. Thus, Flu-NP_366–374/D^b-specific T cells were sorted from the BAL and lung tissue of mice that had recovered from an A/HKx31 infection (35 days postinfection) and tested for direct cytolytic activity against NP_366–374-pulsed L-D^b and control target cells. In several experiments, we were unable to detect direct cytolytic activity in these populations (data not shown). In contrast, Flu-NP_366–374/D^b-specific T cells sorted from in vitro restimulated spleen cultures were strongly cytolytic, confirming that the tetramer used for sorting was not blocking cytotoxic function (16). A similar lack of direct cytotoxic activity was seen with Sen-NP_324–332/K^b-specific T cells isolated from the lungs of mice that had recovered from a Sendai virus infection (40 days postinfection). Given that Ag-specific T cells in the lung were not constitutively cytolytic, we went on to determine whether these cells would produce cytokines ex vivo in a short-term assay. ELISPOT analysis of cells from all tissues identified relatively high frequencies of tetramer⁺ cells producing γ-IFN in response to the appropriate peptides (data not shown).

We also investigated whether Ag-specific CD8⁺ T cells in the lungs could proliferate in response to Ag and acquire cytolytic activity. T cells were isolated from the lung airways, lung tissue, spleen, and MLN of mice that had recovered from an A/HKx31 infection (31 days postinfection). The cells were then labeled with CFSE and restimulated in vitro with the Flu-NP_366–374 peptide Ag and IL-2 for 4 days. Purified lung-derived responder T cells and γ-irradiated stimulator spleen cells were used in these studies to avoid the possible effect of suppressive macrophage populations in the lung (54). As shown in Fig. 5, there was strong proliferation of Flu-NP_366–374/D^b-specific T cells from all four anatomical sites. This response was Flu-NP_366–374 peptide specific inasmuch as no proliferation of Ag specific cells was observed in the absence of peptide (Fig. 5) or with an irrelevant Sen-NP_324–332 peptide (data not shown). We also tested these cultures for cytotoxic activity against Flu-NP_366–374 peptide-pulsed L-D^b target cells. Cells originating from either the lung tissue or airways exhibited strong, specific cytolytic activity that exceeded that observed from spleen and lymph node cultures (Fig. 6). Similar data were obtained with cells taken at later timepoints post A/HKx31 infection, and also following Sendai virus infection (using the Sen-NP_324–332/K^b peptide; data not shown). Thus, although the cells in the lung are not constitutively cytolytic, they have the capacity to proliferate in response to Ag and acquire cytolytic activity.

A key question in these studies is whether the primary pulmonary infection is essential for establishing populations of activated, Ag-specific T cells in the lung. In initial studies, we infected mice i.p. with influenza virus and found that Ag-specific T cell populations were established in the lung in the absence of an overt lung infection (data not shown). However, we could not rule out that there was some virus transfer to the lung under these circumstances. Thus, as an alternative, we performed an adoptive transfer of activated OVA_257–264/K^b-specific transgenic CD8⁺ T cells into adult thymectomized, irradiated, bone marrow-restored hosts. Previous studies have shown that this protocol results in the establishment of memory in the spleen and lymph nodes (44). Ten months after transfer, the presence of donor T cells was analyzed in the lungs and peripheral lymphoid tissues. As shown in Fig. 7, significant numbers of OVA_257–264/K^b-specific T cells could be detected in all tissues including the lung. The distribution of these cells in the lung was similar to that induced by respiratory virus infections in terms of numbers of cells in the lung tissue and airways. In addition, the cells in the lung were highly activated in terms of CD69, CD25, and CD44 expression (Fig. 7), analogous to what had been observed in the infection models. Since the T cells in this experiment had been induced in vitro, these data indicate that neither primary infection in the lung, nor persisting Ag, is required to establish and maintain populations of activated Ag-specific T cells in the lung.

It has been well established that respiratory virus infections induce populations of Ag-specific memory CD8⁺ T cells that persist at high frequencies in the spleen and lymph nodes. The data presented here show that there is also a substantial population of Ag-specific CD8⁺ T cells that persists in the lung after the inflammatory response has resolved and all infectious virus has been cleared (as determined by PCR analysis, egg titration, and T cell hybridoma assays) (46, 52, 53). Immunofluorescence microscopy indicated that these cells were located within the peribronchiolar and perivascular interstitial spaces of the lung. The establishment of Ag-specific T cell populations in the lung appears to be a general characteristic of respiratory virus infections because strikingly similar results were obtained with two independent viral systems, mouse-adapted influenza virus and a natural mouse parainfluenza virus, Sendai virus.

The relationship between memory T cells, recall responses, and cellular immunity in peripheral tissues such as the lungs, skin, and gut is poorly understood. In particular, there is absolutely no information on which populations of memory cells actually mediate a recall response to an infection in the lung in an unmanipulated animal. A particularly puzzling aspect of cell-mediated immunity to respiratory virus infections has been the lack of a direct correlation between the persistence of Ag-specific T cells in the secondary lymphoid organs and protective immunity. For example, in the case of influenza virus infection, protective cellular immunity wanes substantially within 3 mo of infection, despite the fact that Ag-specific memory cells persist at high frequencies in the spleen and lymph nodes for over 1 year after infection (12) (Fig. 1). In this regard, it is interesting that the decline in Ag-specific cell numbers in the lung correlates directly with the loss of protective cellular immunity. Thus, there is a 10- to 20-fold loss in the absolute numbers of Ag-specific T cells in the lung over the first 4 mo postinfection. Moreover, cell-mediated protection never completely disappears and this is consistent with the persistence of low numbers of Ag-specific T cells in the lung. Thus, it is possible that the strength of the recall response to secondary infection depends on the absolute number of Ag-specific T cells already present in the lung. This idea does not necessarily challenge the concept that memory cells resident in the local draining lymph nodes are the key effectors in mediating protective cellular immunity. For example, it is possible that Ag-specific T cells in the lung are necessary for the efficient recruitment of other T cells to the site. In this regard, it is known that T cells secrete chemokines, such as MIP-1α, MIP-1β, and RANTES, which may be involved in accelerating the cellular immune response to a secondary infection. In support of the idea that cells at mucosal sites mediate protective cellular immunity, studies by Bachmann et al. (24) have identified differences between T cell memory in peripheral tissues and systemic sites in terms of protective efficacy and longevity. Thus, memory cells in the spleen appeared be long-lived but play a limited role in protection whereas memory cells in the peripheral tissues appeared be short-lived but offer strong protection. In addition, there appears to be a significant difference in the quality and longevity of cellular immunity elicited by different routes of infection (7, 8, 36, 55, 56). For example, recently published studies by Nguyen et al. (8) indicate that infection of the total respiratory tract induces much stronger cellular immunity than infection by i.p., intranasal (nose only), or i.v. infection. It is possible that the protective efficacy in these cases correlated with the efficiency in which memory T cells were established in the lungs.

From a theoretical standpoint, the advantage of persistent Ag-specific CD8⁺ T cells in the lung is that they are already present at the site of infection when the viral load is still minimal. Thus, even a small number of cells may have a major impact on the course of the infection. In this regard, it should be noted that only about 1–2 × 10⁵ Ag-specific CD8⁺ T cells typically accumulate in the lung at the peak of a primary response to either Sendai or influenza virus infection. Because this peak occurs just after infectious virus has been cleared, it is reasonable to believe that far fewer cells would be necessary for clearing the lower viral loads present at the early phase of a secondary virus infection. Thus, the numbers of Ag-specific T cells present in the lung during the first few months after an infection is likely to be more than sufficient to have a major impact on the control of a secondary infection. We have also observed substantial numbers of CD44⁺/CD69⁺/CD4⁺ T cells in the lung (data not shown) and it is well established that CD4⁺ T cells can play a central role in cellular immunity to respiratory virus infections (7).

Phenotypic analysis of Ag-specific T cells in the lung indicates that they are highly activated in terms of CD69 and CD25 expression. This activated state does not appear to be due to the presence of persistent Ag because viral RNA or processed Ag cannot be detected in the lung beyond 2 wk postinfection. However, it is formally possible that low levels of Ag persist at levels or in forms that cannot be detected by current technology. Interestingly, memory CD8⁺ T cells in the lung do not express the constitutive cytolytic activity characteristic of CD8⁺ T cells at other mucosal sites such as the gut (57). This lack of immediate cytolytic function was analogous to that described in CD8⁺ T cells that develop in lymphocytic choriomeningitis virus-infected mice in the absence of CD4⁺ T cell help (30). However, in contrast to these cells, Ag-specific cells in the lung were able to proliferate and acquire strong cytolytic activity following Ag exposure in vitro. It is interesting to note that similar activated Ag-specific T cells have been described in the brain following intracerebral challenge of mice with influenza virus (58). In this case, activated cells persisted in the brain for at least 320 days after viral challenge in the absence of detectable virus (analyzed by immunohistochemistry and PCR).

We have refrained from referring to the Ag-specific T cells in the lungs as “memory” cells on the basis that they most closely resemble effector cells. Indeed, by most phenotypic criteria, these cells are highly activated, and are distinct from classical resting memory cells. However, memory is defined operationally and, as discussed above, it is possible that persistently activated T cells in the lung play a key role in recall responses and cellular immunity. Thus, it can be argued that functional memory in this system is mediated by this population of cells and that they should be considered memory cells from an operational standpoint. In this regard, it will be crucial to determine whether the population in the lung is a self-sustaining population, or is dependent on the trafficking of activated cells from peripheral lymphoid organs. For example, it is possible that memory cells are continually being activated in the peripheral lymphoid organs and subsequently home to mucosal sites such as the lung. In support of this, several studies have demonstrated that there is a slow turnover of memory cells in the spleen that is sometimes associated with functional activity (17, 28, 59, 60). In addition, Ag-specific CD8⁺ T cells in the lungs appeared to be highly activated in terms of phenotypic markers that are normally associated with TCR engagement. The highest frequency of activated T cells was detected in the lung airways, consistent with the hypothesis that only activated cells can extravasate between the tissues into the airway. Thus, the lung, and possibly other mucosal sites, may act to selectively recruit activated memory cells from the circulating pool. Because the lung is chronically exposed to environmental Ags, a low level of chronic inflammation may serve as the signal for nonspecific recruitment of memory cells. Consistent with this idea, the pattern of persisting memory cells in the lung parenchyma and airways of mice 1 year after re-population with in vitro-generated effectors was very similar to that seen with the virally infected mice. Thus, transferred cells were found in significant numbers in the lung tissue and airways and exhibited the same elevated level of activation marker expression characteristic of cells elicited by virus infection in the lung. These mice received only activated CD8 cells and were not exposed to OVA or viral infection.

Taken together, these data show that Ag-specific CD8⁺ T cells persist in the lungs long after the clearance of a respiratory virus infection. These cells are highly activated and can be induced to proliferate although they do not express constitutive effector function. Importantly, the numbers of activated cells in the lung decline over time providing a potential explanation for the loss of protective cellular immunity in the face of high frequencies of small resting memory cells in the peripheral tissues. Understanding the role of these cells in protective immunity is essential for the development of vaccines designed to emphasize cellular immunity in the lung.

We thank Twala Hogg and Kim Ward for their excellent technical assistance, Brian Helmich and Joyce Reome for the preparation and analysis of the OT-1 “memory” mice, Simon Monard for assistance with the flow cytometry, Jean Brennan and Mike Tighe for help with the immunofluorescence staining, Scottie Adams and Tim Miller of the Trudeau Institute Molecular Biology Core Facility for the generation of tetrameric reagents, and Dr. Marcy Blackman for critically reviewing the manuscript.