NK cells offer a first line of defense against viruses and are considered beneficial to the host during infection. Nevertheless, little is understood regarding the phenotype and function of NK cells in the lung during influenza virus infection. We found that the frequency of NK cells in mouse lung increased during influenza infection, with the majority of a mature phenotype. Cell surface CD107a and intracellular IFN-γ were detected in cells expressing multiple NK-cell receptors in infected lung, suggesting that NK cells were activated during infection. The activating receptor NKp46 was predominantly negative on such cells, possibly as a result of encountering influenza HA. Depletion of NK cells in vivo with anti-asialo GM1 or anti-NK1.1 reduced mortality from influenza infection and surviving mice recovered their body weight. Pathology induced by NK cells was only observed with high, not medium or low-dose influenza infection, indicating that the severity of infection influences NK-cell-mediated pathology. Furthermore, adoptive transfer of NK cells from influenza-infected lung, but not uninfected lung, resulted in more rapid weight loss and increased mortality of influenza-infected mice. Our results indicate that during severe influenza infection of the lung, NK cells have a deleterious impact on the host, promoting mortality.
Natural killer (NK) cells are large granular lymphocytes that mediate innate protection from viruses and tumor cells [1-3]. NK cells directly lyse virally infected cells or tumor cells and produce cytokines and chemokines to attract inflammatory cells to sites of inflammation [3, 4]. Activating and inhibitory receptors expressed by NK cells regulate their functional activity. Activating NK-cell receptors include, but are not limited to, NKG2D, NKp46 (also known as NCR1), FcRγIII, activating Ly49 (in rodents), or activating KIR (in humans) [5, 6]. By contrast, inhibitory Ly49 or KIR and the NKG2A/CD94 heterodimer that recognize MHC class I (MHC-I) ligands or non-MHC specific receptors, such as NKR-P1b and 2B4, maintain NK-cell tolerance [5-7].
Contributions of NK cells toward resistance to viruses can be essential for host health and survival. For example, there is a correlation between humans with NK-cell deficiencies and recurrent and severe infections with varicella zoster and HSVs, respectively [8-10]. Furthermore, the expression of specific activating Ly49 by NK cells can be essential for survival of certain mouse strains from infection by mouse CMV [11, 12]. However, a number of reports demonstrate that NK cells can play an inhibitory role in adaptive immunity [13-16]. In some instances, particularly during lymphocytic choriomeningitis virus (LCMV) infection, this can lead to virus persistence, as well as T-cell-mediated immunopathology [13, 14]. Thus, activities of NK cells can lead to both beneficial and detrimental outcomes from their direct and indirect influences on viral persistence and host immunopathology.
Influenza viruses are one of the most common causes of human respiratory infection and are a major world health concern. Infection with seasonal or pandemic influenza virus strains lead to significant mortality [17, 18]. The most recent pandemic is from swine flu (H1N1) in 2009, a new influenza virus [19, 20]. In 2010, there were over 18 000 deaths worldwide due to this H1N1 strain . Lungs require rapid and effective innate responses to prevent airborne virus infections. On the other hand, immune responses in lung must be tightly regulated to avoid immunopathology and chronic inflammation [22, 23]. Little is understood regarding NK-cell functions and regulatory mechanisms in the lung microenvironment during influenza virus infection. It has been reported that NK-cell depletion or inhibition of NK-cell function in mice can lead to worse morbidity and mortality from influenza virus infection [24-26]. Although this may be the case in mild influenza infection, in this report we demonstrate that NK cells can also be responsible for enhanced morbidity and mortality during more severe influenza infection, which is transferable by NK cells in mice. These results point to the complexity of NK-cell activities and possible regulatory functions of this cell type during influenza infection.
NK-cell kinetics and lung inflammation after influenza virus infection
NK cells not only can destroy virus-infected cells without previous stimulation, but they also can modulate the adaptive immune response [3, 16]. We were interested in determining the nature and function of NK cells in the lung during influenza virus infections. We began by quantifying NK cells in lungs of C57BL/6 mice from day 1 to day 6 postinfection with influenza A/PR8. Compared with mock infection, influenza A/PR8 infection increased the frequency of NK cells in the lung. The percentage of CD3−NKp46+ cells in lung increased fourfold as a result of influenza infection (Fig. 1A). The majority of CD3−NKp46+ cells in influenza-infected lung were NK1.1+ and CD127− (Fig. 1A). Virus-induced NK cells were detected in lung on days 3 and 4 postinfection, whereupon they rapidly declined (Fig. 1B). We also examined splenic NK cells over 6 days postinfection. Lung influenza infection had no influence on the frequency or phenotype of splenic NK cells (data not shown). Despite the rise and fall of NK-cell frequency, there is progressive inflammation in the lung over 6 days of virus infection (Fig. 1C).
Mature NK-cells express inhibitory receptors and an activated phenotype in influenza-infected lung
In addition to NKp46, CD127, and NK1.1 (Figs. 1A and 2A), we characterized the phenotype and lineage markers expressed on NK cells present in influenza-infected lung. The tumor necrosis family member CD27 and integrin CD11b (Mac-1) are markers of the NK-cell lineage . CD11b−CD27+, CD11b+CD27+, and CD11b+CD27− NK cells represent a progression from immature to mature cells with high cytolytic activity, and then to mature cells with limited lytic capability, respectively . At the peak of the NK-cell response to influenza, most lung NK cells are mature CD11b+CD27− cells (Fig. 2B, upper right panel), although a small portion are CD11b+CD27+. NKG2A and Ly49C/I are inhibitory receptors expressed by C57BL/6 NK cells [7, 28]. We found that most NK cells from the lungs of influenza-infected mice express NKG2A and/or Ly49C/I, with a large percentage simultaneously expressing NKG2A and Ly49C/I, or only Ly49C/I, with much smaller percentages expressing only NKG2A, or neither receptor type (Fig. 2B, lower right panel). This pattern of NKG2A and Ly49C/I expression was similar to NK cells in the lung (Fig. 2B, lower left panel) and spleen of uninfected mice (data not shown). Thus, influenza infection had no influence on expression of these inhibitory receptors on lung NK cells.
Elevated CD107a and IFN-γ expression by lung NK cells during influenza virus infection
CD107a is associated with stored intracellular cytolytic granules in NK cells [29, 30]. CD107a appears at the NK-cell surface when they degranulate their cytolytic contents as a result of activation. Thus, NK-cell degranulation activity is estimated by CD107a expression [29, 30]. NK cells also can produce IFN-γ when activated . Furthermore, treatment with IFN-γ can protect mice from death in a NK-cell-dependent manner at an early stage of influenza infection . We purified lymphocytes from influenza-infected lung using Percoll gradients, then stained the cells with anti-CD3 to exclude T cells and identified those which were NK1.1+, CD122hi, 2B4+, and NKp46+, and therefore likely to be NK cells. We found that a small percentage of these cells were positive for CD107a or IFN-γ (Fig. 2C and D), which was slightly more than by these cells in uninfected mice (data not shown). By contrast, a CD3−NK1.1+CD122hi2B4+NKp46− population showed extensive degranulation (over 90% of the cells), and nearly 15% of this population expressed intracellular IFN-γ during influenza infection (Fig. 2C and D). Cells that lacked CD3, expressed the other NK-cell markers, NK1.1, CD122hi, and 2B4, but not NKp46, were not found in any quantity in uninfected mice (data not shown). Downregulation of NKp46 has been described for human NK cells upon encountering influenza virus in vitro, or after in vivo exposure to influenza . Our results suggest that this may also be the case for NKp46 expressed on mouse NK cells isolated from influenza-infected mice. Thus, it is possible that the CD3−NK1.1+CD122hi2B4+NKp46− cells found in influenza-infected lungs are NK cells that have encountered influenza virus and have responded with substantial degranulation and production of IFN-γ.
Improved outcomes for influenza-infected mice by depletion of NK cells
The NK cells in influenza virus infected lung displayed an activated phenotype, suggesting that they play an active not passive role during influenza infection. To investigate the influence of NK cells on host outcome during influenza infection, we treated mice with anti-asialo GM1 to deplete NK cells in vivo prior to and during influenza infection. Anti-asialo GM1 is effective at depletion of NK cells in vivo [34, 35], as confirmed by our flow cytometric analysis of lung and spleen (Fig. 3A). Interestingly, compared with PBS control mice, depletion of NK cells improved the survival rate (Fig. 3B) and recovery of body weight (Fig. 3C) of surviving animals after influenza virus infection. These results suggested that NK cells may exacerbate pathology induced by influenza infection, leading to a worsened outcome.
Virus titer determines the outcome of NK cell depleted influenza-infected mice
Our results (Fig. 3) are contradictory to previous reports [24-26] that found that depletion of NK cells increased mouse morbidity and mortality from influenza infection. In comparing our experiments to those in previous studies, we postulated that influenza virus dose plays a role in determining the influence of NK cells in host morbidity and mortality. To clarify this question, we depleted mice of NK cells in vivo prior to and during infection with different influenza virus titers. Furthermore, anti-NK1.1 was employed as an additional approach to deplete NK cells in these experiments since anti-asialo-GM1 can deplete subsets of cells from other lineages. Flow cytometric analysis confirmed depletion of CD3−NK1.1+ cells in lung and spleen by anti-NK1.1 (Fig. 4A). Depletion of NK cells improved the survival rate and recovery of body weight (Fig. 4B) in high-dose (5 hemagglutination unit (HAU)) influenza infection. Interestingly, the reverse results were found with medium dose (0.5 HAU) influenza infection, that is, depletion of NK cells increased morbidity and mortality in influenza infection (Fig. 4C). In low-dose (0.0625 HAU) influenza infection, compared to PBS control mice, depletion of NK cells did not influence survival rate and recovery of body weight (Fig. 4D). These results indicate that NK cells can be deleterious, beneficial, or inconsequential, depending on the dose of virus that the mice are exposed to.
Transfer of NK cells from influenza-infected lung accelerates weight loss and enhances mortality
Results from NK-cell depletion experiments suggested that NK cells were deleterious during a high-dose pulmonary influenza infection. To further address this issue, we adoptively transferred lung NK cells isolated from high-dose influenza infected or uninfected mice to naive mice, or mice undergoing a primary influenza infection. We purified NK cells from lungs by negative selection before transfer. Flow cytometric analysis confirmed that the purity of adoptively transferred NK cells was greater than 70%, with no contamination by CD8+ T cells in the transferred cells (data not shown). Transferred NK cells were detected in lung and spleen (Fig. 5A). Transferred lung NK cells from influenza-infected mice were not harmful to uninfected recipient mice (v-NK only). By contrast, lung NK cells from high-dose influenza infected mice transferred to recipient mice infected with high-dose influenza virus significantly increased mortality and accelerated body weight loss (Fig. 5B and C). Transfer of lung NK cells from uninfected mice (normal NK) did not alter survival rate or weight loss and recovery kinetics compared to otherwise unmanipulated virus infected recipients. It is possible that influenza virus-induced NK cells enhanced pathology in lung and possibly systemically as well, and either or both contributions may have resulted in the more severe outcome from influenza infection observed. These results are consistent with the NK-cell depletion experiments, and support the conclusion that in high-dose lung influenza infection, NK cells are activated and can enhance mortality.
In the present study, we found that the frequency of NK cells in the mouse lung increased during the course of influenza infection and that depletion of NK cells by administration of anti-asialo GM1 or anti-NK1.1 before and during infection resulted in decreased morbidity and mortality compared to control influenza-infected mice. Not only did a portion of treated animals survive, those surviving animals also experienced rapid recovery to a normal weight range. These findings implicate NK cells in contributing to or exacerbating pathology arising from influenza infection. This contrasts with the described essential function of NK cells in protecting mice from influenza infection, as evidenced by increased morbidity and mortality when NK cells are depleted or rendered less responsive [24-26]. Previous studies have generally used low doses of influenza virus to study NK cell functions and in this case NK cells may contribute significantly to limiting the early propagation of virus. In comparing our experiments with those in previous reports, it appears that virus dose plays a role in determining the overall influence of NK cells in host morbidity and mortality as a consequence of influenza infection. Here, we clarify this issue by showing that increasing the influenza dose from medium to high switches the contribution of NK cells from reducing to enhancing morbidity and mortality. Our results with high-dose influenza infection confirm recent findings by Abdul-Careem et al. , where they observe NK cells contributing to pathology during influenza infection. Unlike our study, Abdul-Careem et al. did neither examine virus dose vis-à-vis the NK-cell contribution to pathology during influenza infection, nor define the importance of this factor, however, the single dose of virus they used may be similar to the high-dose virus level we used in this study, since they obtained similar outcomes . Interestingly, for LCMV infection in mice, it has been demonstrated that the dose of virus greatly affects the influence of NK cells in the immune response to the virus and host outcome. A low dose of LCMV results in viral clearance; a medium dose results in a deleterious NK-cell dependent alteration of T-cell responses, immunopathology, and virus persistence; while with a high dose of virus, NK cells are beneficial by suppressing T cells that would otherwise mediate severe pathology and mortality . It is conceivable that at high influenza dose, the outcome we observed is similar to that seen with infection of mice with a medium dose of LCMV, where there is NK-cell dependent pathology.
The age of mice is another factor affecting host pathology and mortality in the context of influenza infection. This can be seen in comparing the survival curves of the unmanipulated influenza infected control groups in Figures 3 and 5. None of the influenza infected control mice in Figure 3 survived, while approximately 30% of the mice used in Figure 5 did survive the same dose of influenza infection. The mice used in Figures 3 were 4 months old, while those used in Figure 5 were 2 months old. Interestingly, we found that as mice progress from 2 to 4 months of age, NK cells in uninfected lungs can increase from 0.1 to 8.2%, respectively, and in corresponding infected mice could increase to 6.8 and 23.1%, respectively (data not shown). With the frequency of NK cells increasing with age, this could explain why the younger infected control mice survive more frequently (Fig. 5) than their older counterparts (Fig. 3), and is consistent with lung NK cells being detrimental to mice infected with high-dose influenza.
Not only did antibody-mediated reduction of NK cells reduce weight loss and mortality in high-dose influenza infected mice, but adoptively transferring NK cells from influenza-infected mice also exacerbated weight loss and increased mortality in infected mice. To our knowledge, this is the first demonstration of passage of virus-induced NK cell-orchestrated pathology from one animal to another. Also, interestingly, transfer of NK cells from virally infected mice to naïve uninfected mice did not lead to pathology. This may imply that ongoing severe influenza infection in the host may be necessary to sustain expression of effector molecules, expression of relevant NK-cell receptors, and/or induce expression of their ligands on cells of surrounding tissue for NK cells to mediate pathology. The transfer of NK cells from uninfected control mice to virus-infected mice did not enhance weight loss or mortality. This and the preceding discussion may suggest that the contribution of NK cells to pathology is not strictly determined by NK-cell numbers, but possibly whether those NK cells have been adequately exposed to and stimulated by an environment experiencing influenza infection. Our demonstration that cells expressing multiple NK-cell markers in influenza-infected lung largely display an activated phenotype with IFN-γ expression, CD107a at the cell surface, and low cell surface NKp46, is consistent with our adoptive transfer experiments, and suggests that NK cells must be activated to mediate pathology.
The mechanism(s) by which NK cells are exacerbating pathology remains to be elucidated. The NK cells we recovered from lung of influenza-infected mice were mature (CD27loCD11bhi), and many appeared to display an activated phenotype. The expression of cell surface CD107a indicates recent release of cytolytic components including granzymes and perforin [29, 30], suggesting the possibility of direct elimination through cytotoxicity of cells relevant to host protection from virus infection, or perhaps regulatory cells that are capable of restricting pathology. During LCMV infection, NK cells eliminate activated antigen-specific CD4+ T cells, which in turn dampens the CD8+ T-cell response to LCMV . Alternatively, NK cells may indirectly affect lung pathology through the secretion of cytokines and/or chemokines and altering cell interactions and inflammatory responses. The production of IFN-γ by NK cells in lung may be relevant, as IFN-γ is known to limit CD8+ T-cell responses . Another cytokine, IL-22, is known to play a role in mucosal epithelial cell homeostasis and is secreted by NK cells in the lung during influenza infection [38-40], however, neutralization of this cytokine with antibody in vivo was reported to have no effect on weight loss or survival of mice infected with influenza . The receptor(s) responsible for induction of pathology remain to be determined, however, we found that the activating receptor NKp46 was low to negative on many cells expressing multiple NK-cell receptors in influenza-infected lung. Engagement of NKp46, presumably by its ligand influenza HA  might be responsible on its own or in combination with contributions by other activating NK-cell receptors for the activation of NK cells, leading to pathology. Alternatively, or in addition, NK cells can be activated by type I IFN released by DCs as a response to host infection by many diverse viruses , possibly serving as a stimulus for activated NK cell-mediated pathology.
A feature of severe influenza infections in humans leading to mortality, including those by avian H5N1, is massive inflammation in the respiratory tract . Infection of mice with H5N1 or the 1918 pandemic influenza virus [42, 43] results in excessive lung inflammation, as we observe here with high-dose A/PR8 strain infection. NK cells become activated and their numbers reduce in peripheral blood, possibly due to entering the lung, when humans are exposed to seasonal or pandemic strains of influenza . NK cells may assist in orchestrating the excessive infiltration of lung by various cell types during severe influenza infections in addition to or instead of direct cytotoxic functions. High-dose A/PR8 infection in mice may serve as a model for severe influenza infections and the manipulation of NK cells for therapeutic benefit. Partial blocking of NKp46 interactions with influenza HA and/or modulation of Toll-like receptor interactions that lead to NK-cell activation [45-47] may provide an appropriate balance of NK-cell responsiveness during severe influenza infections, such that they are sufficient to mediate protection but not excessive, resulting in pathology.
Our report underscores the complexity of NK-cell influences during the host response to virus infection. Understanding the contributions of NK cells not only to host defense, but also toward pathology during virus infections will aid efforts at manipulation of NK cells for therapeutic efficacy.
Materials and methods
Female C57BL/6 mice at 6–8 weeks of age were purchased from Charles River Laboratories (Kingston, ON, Canada). Experiments were approved by the Animal Welfare and Policy Committee of the University of Alberta (Edmonton, AB, Canada). Housing and handling of mice was in accord with Canadian Council on Animal Care guidelines.
Virus, titers, and infections
Influenza A/PR8 virus was grown in eggs and HAU were determined by hemagglutination assay using chicken red blood cells (Lampire Biological Labs) . For i.n. infection with influenza A/PR8, mice were anesthetized with isoflurane, and then 30 μL of diluted virus (5, 0.5, or 0.0625 HAU) was given. Control mice were given normal egg allantoic fluid i.n. for mock infection. Mice were monitored for weight daily and euthanized when moribund.
Isolation of lung mononuclear cells
Lungs were removed aseptically, and perfused through the right ventricle with 5 mL HBSS to remove peripheral blood cells. To obtain mononuclear cells from lung tissue, the lungs were minced into 2–3 mm sections with scissors and resuspended in DMEM medium supplemented with 10% FBS, 1–2 mg/mL collagenase (Sigma-Aldrich), 50 U/mL DNase (Sigma-Aldrich), HEPES, and antibiotic antimycotic solution (Sigma-Aldrich). The tissues were incubated at 37°C for 60 min with gentle vortexing at 200 rpm. Lung portions were then crushed through 40 μm basket filters and the remaining erythrocytes lysed with lysis buffer (0.15 M NH4Cl, 10 mM KHCO3, 0.1 mM Na2EDTA, pH 7.4) and washed with PBS. The resulting cell suspensions were used for flow cytometric experiments or further cell purification. In some experiments, lymphocytes were purified from lung preparations by Percoll continuous gradient, as previously described , prior to cytometric analysis of NK cells.
Antibodies and flow cytometric analysis
The following purified mouse antigen specific conjugated or unconjugated antibodies: CD16/CD32, CD3-FITC, KLRG1-FITC, NKG2A-FITC, IFNγ-FITC, CD244(2B4)-FITC, Rat IgG2a k Isotype control FITC, NK1.1-allophycocyanin, Mouse IgG2a k Isotype control allophycocyanin, IFN-γ-allophycocyanin, KLRG1-allophycocyanin, CD3e allophycocyanin-eFluor780, CD11b-PE, NK1.1-PE, Ly49C/I-PE, CD107a-PE, NKp46-PE, Rat IgG2a k Isotype control PE, CD107a-PerCP-eFluor710, Rat IgG2a k Isotype control PerCP-eFluor710, CD3-PerCP-eFluor710, NKp46-PerCP-eFluor710, CD27-PerCP-eFluor710, IFN-γ-PerCP-Cy5.5, Rat IgG1 k Isotype control PerCP-Cy5.5, CD122-eFluor450, and Rat IgG2b k Isotype control eFluor450 were purchased from eBioscience (San Diego, CA, USA). CD127-PE-Cy7 was purchased from BD Biosciences. The above-mentioned antibodies were used for FACS analysis in this study. Cells were suspended in buffer comprised of PBS containing 1% FCS plus 0.09% NaN3, followed by incubation with anti-CD16/CD32 mAb and then stained with mAbs specific for cell surface markers for 30 min at 4°C. For intracellular staining, cells were fixed with 4% paraformaldehyde fixative and then stained for 30 min in 0.1% saponin, 0.05% NaN3 in HBSS at room temperature. Events were collected on a BD FACSCanto II, and the data was analyzed using BD FACSDiva software.
NK-cell depletion in vivo
In order to deplete NK cells in mice with influenza infection, mice (4 months old) were i.v. injected with 50 μL anti-asialo GM1  (Wako Chemicals) into the tail vein once every 5 days, starting on day 0. As previously described [50, 51], anti-NK1.1 antibodies were purified from the supernatant of PK136 hybridoma cell culture (American Type Culture Collection, Manassas, VA), and i.v. injected into mice (500 μg/injection) on the same schedule. Control mice were treated with PBS. After depletion treatments, NK cells in lung and spleen were examined by flow cytometry on day 4.
NK-cell adoptive transfer
C57BL/6 mice (2 months old) were i.n. infected with 5 HAU of influenza virus. After 3 days, lung mononuclear cells were isolated from infected mice or uninfected mice, and then the cell suspensions were layered on a Histopaque-1083 gradient (Sigma-Aldrich), and centrifuge at 400 × g for 30 min at room temperature. Subsequently NK cells were purified using a negative selection mouse NK cell enrichment kit (StemCell Technologies), and labeled by CellTrace™ Violet (Invitrogen Corporation). As described previously [52, 53], 2 × 106 NK cells in 0.25 mL PBS were injected i.v. into recipient mice via the tail vein. On the same day, the mice were i.n. infected with 5 HAU of influenza A/PR8 virus. After infection, NK cells from lung and spleen were analyzed by flow cytometry 15 h later. The survival rate and body weight of infected mice were monitored daily.
Two months old C57BL/6 mice were i.n. infected with 5 HAU of influenza virus or normal egg allantoic fluid on day 0. At days 2, 4, and 6 after infection, mice were euthanized and lungs were isolated and fixed in 10% buffered formalin, then embedded in paraffin and sectioned. Specimens were stained with H&E and examined using a Zeiss Axio Imager M1 microscope equipped with an AxioCam HRc camera under control of AxioVision 4 software (Carl Zeiss Canada Ltd.).
GraphPad Prism 4.00 (GraphPad Software, Inc., San Diego, CA, USA) was used for all analyses. Differences among experimental groups were assessed by one-way ANOVA followed by Tukey multiple comparison test. Unpaired t-test (two-tailed) was used to compare pairs of groups. Survival curves were assessed by survival analysis in Prism. Values were reported as the mean ± SEM.
This work was supported by operating grants from the Canadian Institutes for Health Research (to K.P.K.). We thank Suellen Lamb, Dr. L. Tyrrell Laboratory, University of Alberta for making histologic sections and performing hematoxylin and eosin staining. We thank Donger Gong for her technical support.
Conflict of interest
The authors declare no financial or commercial conflict of interest.