Suppression of Cytotoxic T Cell Functions and Decreased Levels of Tissue Resident Memory T cell During H5N1 infection.

Influenza viruses cause upper respiratory tract infections in humans. In healthy adults, seasonal influenza virus infections result in mild disease. Occasionally, influenza viruses endemic in domestic birds can cause severe and fatal disease even in healthy individuals. In avian influenza virus-infected patients, the host immune system is activated in an uncontrolled manner and is unable to control infection in a timely fashion. In this study, we investigated why the immune system fails to effectively control a modified form of avian influenza virus. Our studies show that T cell functions important for clearing virally infected cells are impaired by higher negative regulatory signals during modified avian influenza virus infection. In addition, memory T cell numbers were decreased in modified avian influenza virus-infected mice. Our studies provide a possible mechanism for the severe and prolonged disease associated with avian influenza virus infections in humans.

I nfluenza A viruses, members of the Orthomyxovirus family, cause upper respiratory infections in humans (1). Infections by seasonal influenza A virus strains (H1N1 and H3N2) are mostly self-limiting in healthy adults; however, seasonal infections can be severe in young children and the elderly (2, 3). In addition to humans, influenza viruses can infect a variety of zoonotic species, including domestic poultry, pigs, horses, seals, and waterfowl (4)(5)(6). Occasionally, influenza virus strains circulating in zoonotic reservoirs can cross the species barrier and cause infections in humans. Unlike seasonal H1N1 and H3N2 strains, infections with avian influenza viruses such as H5N1 and H7N9 are often severe in all age groups and cause extensive alveolar damage, vascular leakage, and increased infiltration of inflammatory cells in the lungs. The virulent nature of avian influenza viruses has been attributed to both viral and host determinants; while the viral determinants of virulence are well defined, the contribution of host responses to disease severity remain to be elucidated.
The H5N1 strain of avian influenza virus was first detected in humans during a domestic poultry outbreak in Hong Kong in 1997 (7,8). Despite considerable efforts for containment, H5N1 strains have spread globally and are now endemic in domestic poultry on several continents. Over the past 20 years, H5N1 viruses from infected domestic poultry have crossed the species barrier, causing severe and often fatal infections in humans, with mortality rates as high as 60% (9). Many of the viral components critical for the enhanced virulence of H5N1 have been identified through the generation of recombinant and/or reassortant viruses (10)(11)(12). Prior studies have shown that the multibasic cleavage site (MBS) in the viral hemagglutinin of H5N1 facilitates higher viral replication and mediates extrapulmonary spread (13)(14)(15). In addition, our group has recently demonstrated that the endothelial cell tropism of H5N1 contributes to barrier disruption, microvascular leakage, and subsequent mortality (12). Moreover, polymorphisms that increase viral replication have been identified in the viral polymerase subunits of H5N1 strains (16)(17)(18)(19)(20). Together, these studies have helped to define the viral components that are responsible for the enhanced virulence of H5N1.
Apart from viral determinants, overt and uncontrolled activation of the innate immune responses also contribute to the disease severity associated with H5N1 infection (21,22). Histological analyses of lungs from fatal H5N1 cases demonstrate severe immunopathology, as evidenced by excessive infiltration of immune cells into the lungs and greater numbers of viral antigen-positive cells in the lungs (23,24). In corroboration with these studies, H5N1 viruses have been shown to induce higher dendritic cell (DC) activation and increase cytokine production compared with H1N1 viruses (25). Moreover, studies with H5N1 strains in animal models demonstrate hyperactivation of resident immune cells in the lungs and a consequent upsurge in cytokine levels (26,27). As such, these heightened proinflammatory responses result in the excessive recruitment of neutrophils and inflammatory monocytes into the lungs, correlating with severe disease (24). Despite robust activation of innate immune responses against H5N1 infection, higher and prolonged virus replication can be detected in the lungs of infected individuals, suggesting a possible dysregulation of adaptive immune responses (28).
We have previously demonstrated that appropriate activation of respiratory DCs is required for effective T cell responses against a mouse-adapted H1N1 strain (29). Here, we sought to determine if excessive activation of innate immune cells during avian H5N1 infection impairs subsequent adaptive T cell responses. In order to investigate the immune responses against H5N1 compared with a mouse-adapted H1N1 strain, we generated a closely matched recombinant H5N1 virus carrying the 6 internal genes of H1N1 (H5N1 (2:6)). Our studies demonstrated that H5N1 (2:6) infection in mice induced higher lung DC activation and promoted increased migration of lung DCs to the draining lymph nodes, resulting in increased numbers of virus-specific CD8 ϩ and CD4 ϩ T cells in the lungs compared with H1N1-infected mice. Despite greater numbers of virus-specific T cells, we observed delayed clearance of H5N1 from the lungs, which correlated with higher PD-1 expression and increased production of the anti-observed in the MLN of H5N1 (2:6)-infected mice resulted in enhanced T cell responses and viral clearance. To evaluate primary T cell responses, C57BL/6 mice were infected with 100 PFU of H5N1 (2:6) or H1N1 and T cell responses were measured on day 8 p.i. by tetramer staining and by monitoring for cytokine production upon ex vivo stimulation. CD8 ϩ T cell responses against H5N1 and H1N1 viruses encompass many epitopes, and here, we investigated the T cell response against immunodominant    (2:6) and H1N1. C57BL/6 mice were infected with 100 PFU of H1N1 or H5N1 (2:6), and CD86 expression was measured flow cytometry. (E) Representative histograms of CD86 expression on lung DC subsets. (F) Quantification for panel E shown as MFI. (G) Comparison of Mx1, ISG15, and IFN-␤ expressions between H5N1 (2:6)-and H1N1-infected lungs. Total RNA was extracted from lung homogenates of infected mice isolated on day 4 p.i. and subjected to reverse transcriptase quantitative PCR (qRT-PCR) analysis. The values are expressed as mean Ϯ SD. *, **, and *** denote significance of Ͻ0.05, Ͻ0.01, and Ͻ0.001, respectively. Data are representative of at least three independent experiments. epitopes (polymerase acidic [PA] and nucleocapsid protein [NP]) elicited in H5N1 (2:6) and H1N1 virus infections. Using tetramers specific for viral NP or PA, we observed increased frequencies of both NP and PA tetramer-positive CD8 ϩ T cells in H5N1 (2:6)-infected mice compared with H1N1-infected mice ( Fig. 3A and B). The absolute numbers of NP and PA tetramer-positive CD8 ϩ T cells were also greater in H5N1(2:6)infected mice than in H1N1-infected mice (Fig. 3C). In addition, ex vivo stimulation with X-31 (H3N2) virus or viral peptides showed increased frequencies of interferon gamma (IFN-␥)and granzyme B (GrB)-producing CD8 ϩ T cells in H5N1 (2:6)-infected mice compared with H1N1-infected mice (Fig. 3D). Moreover, H5N1 (2:6)-infected mice showed increased frequencies of IFN-␥-and GrB-producing CD4 ϩ T cells compared with H1N1-infected mice (Fig. 3E). These results demonstrate that hyperactivated lung DCs promote robust activation of virus-specific T cell responses in the lung.
In the mouse model of influenza virus, innate immune cells restrict viral replication prior to the establishment of adaptive T cell responses. However, after day 6 p.i., T cells primed in the MLN migrate to the lungs and participate in the clearance of virusinfected cells. Therefore, we determined if the greater numbers of virus-specific T cells observed in H5N1 (2:6)-infected mice resulted in efficient viral clearance in the lungs. C57BL/6 mice were infected with 100 PFU of H5N1 (2:6) or H1N1; viral loads in the lungs were measured by plaque assay at various days p.i. Prior to and including day 6 p.i., we observed similar viral loads in the lungs of both groups of infected mice, suggesting that both viruses replicate to similar levels (Fig. 3F). However, on day 8 and day 9 p.i., we observed greater viral loads (ϳ5-to 10-fold) in the lungs of H5N1 (2:6)-infected mice than that of H1N1-infected mice. These results demonstrate that, despite the presence of more virus-specific T cells in the lungs, viral clearance was delayed in H5N1 (2:6)-infected mice.
To understand the basis for the delayed clearance of H5N1 (2:6) in the lungs, we evaluated the functionality of T cells by in vitro T cell killing assay. In this assay, T cells isolated from H5N1 (2:6) or H1N1-infected mice were cocultured with CFSE-labeled splenocytes pulsed with the NP peptide, and the amount of target cell death was determined by quantification of 7-aminoactinomycin D (7-AAD)-positive splenocytes. Interestingly, we observed more splenocyte death in cocultures containing T cells from H5N1 (2:6)-infected mice than in cocultures containing T cells from H1N1-infected mice ( Fig. 3G and H). Next, we performed an in vivo killing assay with peptide-pulsed splenocytes. Splenocytes were labeled with either low CFSE or high CFSE and pulsed with influenza virus NP peptide or control peptide, respectively. Splenocytes were adoptively transferred into mice previously infected with either H5N1(2:6) or H1N1 (day 8 p.i.) ( Fig. 3I and J) and 8-h post-adoptive transfer mice splenocytes were analyzed for CFSEϩ cells. Our data demonstrate that cytotoxic T cells from H5N1 (2:6)-infected mice can effectively kill peptide-pulsed splenocytes both in vitro and in vivo.
Cytotoxic T cells from H5N1 (2:6)-infected mice show higher expression of PD-1 and IL-10. T cell functions can be modulated by stimulatory as well as inhibitory signals. Prior studies demonstrate that during influenza virus infection, T cell functions can be suppressed by PD-1/PD-L1 interactions and by the anti-inflammatory cytokine IL-10 (30-32). In addition, PD-1 has shown to be upregulated in T cells in response to direct activation of T cell receptor (TCR). Although the T cells isolated from H5N1 (2:6)-infected mice were efficient in killing peptide-pulsed splenocytes, we observed delayed viral clearance in the lungs ( Fig. 3F and G). Thus, we investigated if the T cell functionality was suppressed in vivo through PD-1/PD-L1 interactions by measuring the expression of PD-1/PD-L1 by flow cytometry. We observed significantly higher levels of PD-1 on CD8 ϩ T cells isolated from H5N1 (2:6)-infected mice than that from H1N1infected mice ( Fig. 4A and B). Next, we analyzed different cellular compartments in the lungs for PD-L1 expression and observed significantly higher levels of PD-L1 on inflammatory monocytes (CD11bϩ Ly6C hi Ly6G Ϫ ) isolated from H5N1 (2:6)-infected mice than that from H1N1-infected mice (Fig. 4C). In addition, we observed increased numbers of inflammatory monocytes in the H5N1(2:6)-infected mice group (Fig. 4D). However, the levels of PD-L1 on other cellular compartments in the lungs, including inflammatory DCs were similar between the two groups. Next, we measured IL-10 production in T cells isolated from infected mice to determine the possibility of IL-10-mediated suppression of T cell functions. T cells isolated from H5N1 (2:6) or H1N1-infected mice on day 8 p.i. were cocultured with DCs pulsed with the major histocompatibility complex class I (MHC-I) or MHC-II peptide or infected with X-31 (H3N2) virus, and the production of IFN-␥ and IL-10 in T cells was measured by flow cytometry. We observed an increased production of IFN-␥ and IL-10 in both CD8 ϩ and CD4 ϩ T cells isolated from H5N1 (2:6)-infected mice compared with H1N1-infected mice ( Fig. 4E and F). Taken together, these results demonstrate that H5N1 (2:6) infection results in higher expression of inhibitory signals such as PD-1 and IL-10 by T cells, which likely suppress cytotoxic T cell functions in vivo.

H5N1-infected mice show decreased numbers of memory T cells in the lung parenchyma.
Upon clearance of viral infection, a portion of virus-specific T cells differentiate into tissue-resident memory T (T RM ) cells, which play an important role in providing heterosubtypic immunity against subsequent influenza virus infections (33). As we observed higher upregulation of inhibitory signals (PD-1 and IL-10) on T cells from H5N1 (2:6)-infected mice, we investigated if T RM cell responses were also impaired. Previously, T RM cells in lungs post-influenza virus infection have been characterized with the expression of CD69 and CD103 on day 30 p.i. (34). We analyzed the lung parenchyma for memory T cells that exhibit the T RM cell phenotype (CD69 ϩ CD44 ϩ CD103 ϩ ) by flow cytometry (35,36). Circulating T cells were excluded by intravenous injection of labeled anti-CD8␤ antibody prior to euthanizing mice and excluding this population from analysis (gating strategy shown in Fig. 5A). We observed fewer numbers of NP tetramer-positive CD8 ϩ T RM cells in the lungs of H5N1 (2:6)-infected mice than that in H1N1-infected mice (Fig. 5B). Similarly, the numbers of CD4 ϩ T RM cells were fewer in H5N1 (2:6)-infected mice than in H1N1-infected mice (Fig. 5C). These results demonstrate that H5N1 (2:6) infection results in decreased differentiation of lung-resident memory T cells. Next, to determine if the fewer numbers of tissueresident memory cells affect protection from future challenge, C57BL/6 mice were infected with 50 PFU of H1N1 or H5N1(2:6) virus and were subsequently challenged with a heterologous H3N2 strain (X-31), a reassortant strain that share 6 internal genes with H1N1 and H5N1(2:6) viruses (Fig. 5D). We did not observe significant differences in weight loss between H1N1-and H5N1 (2:6)-infected groups upon lethal challenge with the H3N2 (X-31) strain. These data suggest that the lowered levels of memory T cells in H5N1 (2:6) infection do not impact protection against heterologous strains.

DISCUSSION
Infections with avian H5N1 influenza virus induce higher innate immune responses than human H1N1 viruses (21,37). However, due to inherent differences in replication levels, it is difficult to discern if this hyperactivation of innate immune responses against H5N1 is due to higher viral replication in the lungs. To overcome this caveat, we generated an H5N1 strain sharing the 6 internal genes of H1N1 (H5N1 (2:6)) and observed that the HA and NA of H5N1 can induce higher activation of lung DCs. As such, this heightened stimulation of lung DCs by H5N1 (2:6) resulted in increased migration of DCs to the MLN and induced robust T cell responses compared with H1N1 virus. Interestingly, despite the greater numbers of virus-specific T cells in the lungs, we observed delayed clearance of H5N1 (2:6) from the lungs of infected mice. This delayed viral clearance correlated with increased levels of PD-1 expression and IL-10 production by CD8 ϩ T cells, which likely suppress cytotoxic T cell functions in vivo. Importantly, H5N1 (2:6) infection resulted in decreased numbers of tissue-resident memory T cells compared with H1N1 infection. Taken together, our studies demonstrate that hyperactivation of the innate immune system by H5N1 (2:6) results in suppression T cell functions, delayed viral clearance, and decreased numbers of tissue-resident memory T cells.
Unlike seasonal influenza viruses, avian H5N1 influenza viruses can cause severe and often fatal disease in healthy individuals (38,39). H5N1 infection induces uncontrolled activation of the host immune system, with heightened cytokine levels in the lungs as well as massive infiltration of neutrophils, inflammatory monocytes, and inflammatory tumor necrosis factor alpha/inducible nitric oxide synthase (TNF-␣/iNOS)-producing (Tip) DCs (40). These infiltrating cells have been implicated in the enhanced virulence of avian H5N1 influenza viruses (26,27,40). Moreover, ex vivo studies show that H5N1 viruses induce higher human DC activation than H1N1 viruses (25). Similarly, our studies with H5N1 (2:6) virus demonstrated increased activation of murine lung DC compared with H1N1 virus, further suggesting that the H5N1 HA/NA are sufficient for higher activation of innate immune cells in the lungs (Fig. 1E and F). This increased activation of lung DCs in H5N1 (2:6)-infected mice was not due to differences in viral replication between strains, as both the reassortant H5N1 (2:6) and H1N1 (PR8) strains showed similar levels of viral replication on days 2, 4, and 6 p.i. (Fig. 3F). These results corroborate prior ex vivo studies which indicate that viruses with different HA subtypes can differentially activate primary DCs and macrophages (41). However, the consequence of higher DC activation in vivo to subsequent adaptive immune responses was previously unknown. Our studies demonstrated that H5N1 (2:6) infection stimulated increased migration and accumulation of DCs in the MLN, resulting in robust T cell responses in the lungs (Fig. 2C to E). Moreover, we observed increased frequencies of cytokineproducing T cells in H5N1 (2:6)-infected mice compared with H1N1-infected mice (Fig.  3). Together, these data demonstrate that hyperactivation of lung DCs results in increased numbers of virus-specific T cells in the lungs of H5N1 (2:6)-infected mice.
Prior studies demonstrate that the magnitude and the quality of T cell responses determine the efficiency of viral clearance (42). Previously, we demonstrated that mice deficient in retinoic acid inducible gene I (RIG-I) or mitochondrial antiviral signaling protein (MAVS) mounted poor T cell responses against influenza virus, as evidenced by decreased numbers of polyfunctional T cells and delayed viral clearance in the lungs (29). In contrast, despite mounting robust T cell responses, H5N1 (2:6)-infected mice showed delayed viral clearance in the lungs compared with H1N1-infected mice (Fig.  3F). This delayed viral clearance in H5N1 (2:6)-infected mice was likely due to active suppression of cytotoxic T cell functions in vivo, as T cells isolated from H5N1 (2:6)infected mice showed efficient cytotoxic activity against NP peptide-pulsed splenocytes (Fig. 3G). In corroboration, we observed higher levels of inhibitory signals (PD-1 and IL-10) that likely suppress cytotoxic T cell functions in vivo and delay viral clearance ( Fig.  4A and 3F). In our in vivo killing assays, H5N1 (2:6)-infected mice showed robust killing of viral peptide-loaded splenocytes, suggesting that inhibition of T cells may occur by direct suppression by cell-to-cell contact rather than by the presence of suppressive cytokine milieu. Prior studies indicate that infection with the lethal mouse-adapted PR8 strain (H1N1) resulted in higher PD-1 expression on T cells than the less-virulent X-31 (H3N2) reassortant strain (30). Interestingly, our studies show that infection with H5N1 reassortant (2:6) induced higher PD-1 expression than PR8 (H1N1) (Fig. 4A). As H5N1 viruses have been shown to have broad tissue tropism, it is possible that the increased PD-1 expression observed in H5N1 (2:6)-infected mice is likely due to antigen persistence and/or prolonged stimulation of T cells. PD-1 interactions with PD-L1 have been demonstrated to suppress cytotoxic CD8 ϩ T cell functions (43)(44)(45). PD-L1 expression is induced during viral infection on a variety of cell types, including monocytes, DCs, macrophages, and epithelial cells (44)(45)(46)(47). In our analysis of cell types expressing PD-L1, we observed higher PD-L1 expression on Ly6C hi inflammatory monocytes isolated from H5N1 (2:6)-infected mice than that from H1N1-infected mice (Fig. 4C). In addition, the numbers of inflammatory monocytes were higher in H5N1(2:6)-infected mice than in H1N1-infected mice. It should be noted that PD-L1 expression was observed on others cell types as well; yet, there was no significant difference in PD-L1 levels between the two groups (data shown for inflammatory DCs) (Fig. 4C). In a prior study, anti-PD-L1 treatment of PR8-infected mice showed increased virus-specific T cells and decreased viral titers (30); however, anti-PD-L1 treatment did not alter disease outcome, suggesting that there may be additional mechanisms for suppression of T cell functions. In agreement, we observed an increased expression of the anti-inflammatory cytokine IL-10 in T cells isolated from H5N1 (2:6)-infected mice compared with H1N1-infected mice (Fig. 4C to E). Taken together, our data suggest that higher levels of IL-10 production and PD-1/PD-L1-mediated inhibition likely contribute to suppression of T cell functions and consequently results in delayed clearance of H5N1 (2:6) in the lungs.
Upon viral clearance in the lungs, a portion of virus-specific T cells differentiate into tissue-resident memory T cells, and these T RM cells are critical for providing heterosubtypic immunity (36,48). Interestingly, we observed decreased numbers of T RM cells in H5N1 (2:6)-infected mice compared with H1N1-infected mice (Fig. 5). It should be noted that despite the decreased number of T RM cells, we did not observe significant differences in protection against challenge from a heterologous H3N2 strain. Our future studies will determine if fewer numbers of T RM cells are due to defects in differentiation versus maintenance of T RM cells. Prior studies indicate that transforming growth factor-␤ (TGF-␤) promotes maturation of T RM by inducing the upregulation of CD103 expression (36,(49)(50)(51). Coincidently, influenza viral neuraminidase (NA) can convert latent TGF-␤ into mature TGF-␤; however, the NA of H5N1 is unable to activate TGF-␤ both in vitro and in vivo (52,53). It is possible that the decreased numbers of T RM cells in H5N1 (2:6)-infected mice may result from lowered levels of TGF-␤ activation by viral NA. Apart from TGF-␤, interleukin-33 (IL-33) and tumor necrosis factor (TNF) are also known to induce T RM cell-like phenotypes (CD69 ϩ CD103 ϩ ) (36,(49)(50)(51). Moreover, homeostatic cytokine IL-15 is required for T RM cell differentiation and survival (36). Thus, it is also possible that H5N1 (2:6) infections may alter the levels of other cytokines that are critical for the generation and maintenance of T RM cells. Alternatively, sustained inflammation during H5N1 (2:6) infection may negatively regulate T RM cell differentiation due to higher levels of IFN-␤ and IL-12 (54). Further studies are needed to determine if decreased TGF-␤ levels or sustained higher inflammation in the lungs of H5N1 (2:6)-infected mice are responsible for the inefficient differentiation of the T RM population.
In conclusion, our study demonstrates that hyperactivation of innate immune cells by H5N1 (2:6) dampens T cell responses and delays viral clearance in the lungs. This is likely due to higher expression of the inhibitory molecule PD-1 on T cells as well as higher production of the anti-inflammatory cytokine IL-10 by T cells in H5N1 (2:6)infected mice. As such, our studies show that suppression of T cell responses may contribute to the protracted viral replication and prolonged illness associated with avian influenza virus infection in humans.

MATERIALS AND METHODS
Ethics statement. All studies were performed in accordance with the principles described by the Animal Welfare Act and the National Institutes of Health guidelines for the care and use of laboratory animals in biomedical research. The protocols for performing murine studies were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Chicago.
Viruses. The generation of H1N1-GFP (A/Puerto Rico/8/1934) was described earlier (55). H5N1-GFP (A/Vietnam/1203/2004; low pathogenic without the multibasic site in HA), which contains a GFP reporter in the NS segment, was generated following a similar protocol (55). H5N1 (2:6) (A/Vietnam/1203/2004; low pathogenic without the multibasic site in HA), which contains the 6 internal genes from the PR8 strain, was rescued using standard reverse genetics techniques (55,56). Briefly, 0.5 g of each of the six pDZ plasmids representing PB2, PB1, PA, NP, NS, and M from A/Puerto Rico/8/1934 (PR8) and two pPol-I plasmids representing the HA (low pathogenic) and NA segments of H5N1 were transfected into a cell mixture containing 293T-MDCK using Lipofectamine 2000 (Invitrogen). After 48 h, 200 l of the rescue supernatants was used to infect fresh MDCK cells seeded in 6-well plates. The successful rescue of recombinant viruses was confirmed by performing a hemagglutination assay with chicken red blood cells. After plaque purification, the recombinant viruses were amplified in 10-day-old specific-pathogenfree eggs (Charles River). Viral titers were determined by plaque assay in MDCK cells using standard techniques.
Mice infection. C57BL/6 mice were purchased from Jackson Laboratory and bred in specificpathogen-free (SPF) facilities maintained by the University of Chicago Animal Resource Centre. All experiments were performed with gender-matched mice of 6 to 8 weeks of age. For influenza virus infections, mice were anesthetized with ketamine/xylazine (intraperitoneally [i.p.]; 80/10 mg/kg) and infected intranasally with the indicated dose of virus diluted in 25 l of phosphate-buffered saline (PBS).
Reverse transcriptase quantitative PCR analysis. Total RNA from lung tissue was extracted using TRIzol (Life technologies) following the manufacturer's instructions, and cDNA was synthesized with SuperScript II using oligo dT primers (Roche Diagnostics). Quantitative PCR was performed using previously described gene-specific primers in an ABI 7300 real-time PCR system with SYBR green master mix (Invitrogen) (12).
DC and T cell assays. Bone-marrow-derived dendritic cells (BMDCs) were generated from C57BL/6 mice, and T cell restimulation experiments were performed as previously described (57,58). Briefly, BMDCs were infected with X-31 (H3N2) at a multiplicity of infection (MOI) of 0.5 for 5 h, washed with PBS 3 times to remove unbound virus, and resuspended in Iscove's modified Dulbecco's medium (IMDM) with 10% FBS (Invitrogen). T cells from the lungs of naive or infected mice (day 8 p.i.) were enriched using the Pan T cell isolation kit II (Miltenyi Biotec) and cocultured with infected BMDCs at a ratio of 10:1 for 2 to 3 h, followed by the addition of brefeldin A (5 g/ml; eBiosciences). The cells were further incubated for an additional 8 to 10 h at 37°C. Ex vivo peptide stimulation studies were performed using MHC-I NP 366-374 (ASNENMETM) or MHC-II-restricted NP 311-325 (QVYSLIRPNENPAHK) peptides. The cells were first stained for cell surface markers as described above, followed by intracellular staining for cytokines. For intracellular staining, cells were incubated in permeabilization and fixation buffer (BD Pharmingen) for 45 minutes, followed by 2 washes in a PBS buffer containing 1% FBS and 0.5% saponin (Sigma, St. Louis, MO). Intracellular staining for anti-IFN-␥ (2 g/ml; clone XMG1.2; BioLegend), anti-granzyme B (2 g/ml; clone GB11; BioLegend), and IL-10 (2 g/ml; clone JES5-16E3; BioLegend) was performed on ice for 30 minutes.
For T cell tetramer staining, lymphocytes from the lungs of influenza virus-infected mice were enriched using Ficoll-Hypaque (GE Healthcare Life Sciences) density gradient and stained with H-2D b restricted tetramers conjugated to fluorophore R-phycoerythrin (PE) (NP 366-374 ASNENMETM or PA 224-233 SSLENFRAYV).
In vivo killing assay. Single-cell suspension was prepared from mice spleen, and the cells were pulsed either with 1 M NP 366-374 peptide or OT-1 peptide (Ova 257-264 ) for 1 h and labeled with 1 M CFSE (CFSE low ) or 5 M CFSE (CFSE high ), respectively, following manufacturer's instructions (Life Technologies). A mixture of 2 ϫ 10 6 CFSE low and CFSE high splenocytes were intravenously injected to gendermatched naive mice or the mice which had been intranasally infected 8 days ago with H1N1 or H5N1 (2:6) virus. After 5 h of injection, mice splenocytes were analyzed for CFSE-positive cells by flow cytometry. Percent killing was determined using the following equation: % specific lysis ϭ 100 Ϫ [100 ϫ (% CFSE low infected mouse/% CFSE high infected mouse)/(% CFSE low naive mouse/% CFSE high naive mouse).
Statistical analysis. Data were analyzed using Prism GraphPad software, and statistical significance was determined by one-way analysis of variance (ANOVA) or the unpaired Student's t test. *, **, and *** denote significance of Ͻ0.05, Ͻ0.01, and Ͻ0.001, respectively; ns denotes not significant.