Influenza A Virus Infection in Pigs Attracts Multifunctional and Cross-Reactive T Cells to the Lung

ABSTRACT Pigs are natural hosts for influenza A viruses and play a critical role in influenza epidemiology. However, little is known about their influenza-evoked T-cell response. We performed a thorough analysis of both the local and systemic T-cell response in influenza virus-infected pigs, addressing kinetics and phenotype as well as multifunctionality (gamma interferon [IFN-γ], tumor necrosis factor alpha [TNF-α], and interleukin-2 [IL-2]) and cross-reactivity. A total of 31 pigs were intratracheally infected with an H1N2 swine influenza A virus (FLUAVsw) and consecutively euthanized. Lungs, tracheobronchial lymph nodes, and blood were sampled during the first 15 days postinfection (p.i.) and at 6 weeks p.i. Ex vivo flow cytometry of lung lymphocytes revealed an increase in proliferating (Ki-67+) CD8+ T cells with an early effector phenotype (perforin+ CD27+) at day 6 p.i. Low frequencies of influenza virus-specific IFN-γ-producing CD4+ and CD8+ T cells could be detected in the lung as early as 4 days p.i. On consecutive days, influenza virus-specific CD4+ and CD8+ T cells produced mainly IFN-γ and/or TNF-α, reaching peak frequencies around day 9 p.i., which were up to 30-fold higher in the lung than in tracheobronchial lymph nodes or blood. At 6 weeks p.i., CD4+ and CD8+ memory T cells had accumulated in lung tissue. These cells showed diverse cytokine profiles and in vitro reactivity against heterologous influenza virus strains, all of which supports their potential to combat heterologous influenza virus infections in pigs. IMPORTANCE Pigs not only are a suitable large-animal model for human influenza virus infection and vaccine development but also play a central role in the emergence of new pandemic strains. Although promising candidate universal vaccines are tested in pigs and local T cells are the major correlate of heterologous control, detailed and targeted analyses of T-cell responses at the site of infection are scarce. With the present study, we provide the first detailed characterization of magnitude, kinetics, and phenotype of specific T cells recruited to the lungs of influenza virus-infected pigs, and we could demonstrate multifunctionality, cross-reactivity, and memory formation of these cells. This, and ensuing work in the pig, will strengthen the position of this species as a large-animal model for human influenza virus infection and will immediately benefit vaccine development for improved control of influenza virus infections in pigs.

I n 2016, almost 100 years after the devastating 1918 influenza pandemic in humans, influenza A viruses remain a challenge for vaccine development. Antigenic drift and reassortment of influenza virus genomes enable evasion from serological herd immunity, resulting in annual epidemics and unpredictable pandemic outbreaks (1). Reassortment often occurs in pigs, which are susceptible to both avian and human-adapted influenza viruses (2), and these animals have therefore been suggested as mixing vessels, providing ideal conditions for the production of new pandemic strains (3,4). Bidirectional influenza virus transmissions between humans and pigs are known to occur frequently (5)(6)(7)(8)(9). The pandemic outbreak of "swine flu" in 2009 demonstrated how easily reassorted strains of pig origin can "jump" to a naive human population (10) and led to calls for increased surveillance and improved control of influenza in pigs (11)(12)(13).
Current influenza vaccines for both pigs and humans elicit primarily strain-specific humoral immunity, failing to protect against strains carrying drift variants or reassorted genome segments of hemagglutinin (HA). In the quest to develop broadly protective vaccines, T cells have increasingly gained attention, as they are able to recognize internal epitopes that are highly con-served across influenza virus subtypes (14). The important role of T cells in the clearance of influenza in mice (15,16) and their cross-reactive potential (17,18) have long been known. More recent mouse studies have provided evidence that memory T cells in the lung are key to protect against influenza virus infection (19)(20)(21)(22). Evidence for a protective role of T cells also comes from nonhuman primate models (23,24), which more closely approximate human infection. In humans, preexisting influenza virus-reactive T cells and the rapid onset of influenza virus-specific T-cell responses, as measured in blood, could be correlated with reduced symptom scores and rapid recovery from infection, respectively (25)(26)(27). Local lung responses are difficult to assess in humans, but influenza virus-reactive T cells with a tissue-resident memory phenotype could be detected in human lungs obtained by lobectomy (28,29) and from organ donors (30).
Despite the serious zoonotic threat posed by influenza virusinfected pigs and their suitability as large-animal models for human vaccine development (31), in-depth data on porcine T-cell immunity to influenza virus are scarce. Many studies support the involvement of T cells in porcine influenza virus infection (32)(33)(34)(35)(36)(37)(38)(39)(40)(41), but only recently we reported on the first comprehensive study of T-cell kinetics, phenotype, and quality (42). We were able to demonstrate influenza virus specificity, multifunctionality, and memory responses of blood-derived CD4 ϩ and CD8 ϩ T cells in pigs. These longitudinal data from peripheral blood clearly supported an important role of T cells in the porcine immune response against influenza virus infection, though we emphasized the urgent need for T-cell analyses at the site of infection. With the present study, we provide the first thorough characterization of the local T-cell response in the porcine influenza virus-infected lung, addressing kinetics and phenotype, as well as cytokine profile, memory formation, and cross-reactivity.

Animal infection experiments.
In three consecutive animal infection experiments, a total of 31 10-week-old pigs were administered swine influenza A virus (H1N2 FLUAVsw) intratracheally. Another 31 pigs underwent the same procedure but received phosphate-buffered saline (PBS) instead ( Fig. 1). Pigs ([Landrace ϫ Large White] ϫ Pietrain) were obtained from a conventional breeding farm in the federal state of Lower Austria. Breeding sows on this farm had repeatedly tested negative for anti-influenza A virus nucleoprotein antibodies (ID Screen Influenza A Antibody Competition Multispecies enzyme-linked immunosorbent assay [ELISA]; IDvet, Grabels, France) as well as antibodies against porcine reproductive and respiratory syndrome virus (PRRSV) (tested by PRRS X3 ELISA; Idexx, Ludwigsburg, Germany). The study pigs also tested negative for anti-influenza A virus nucleoprotein antibodies prior to experimental infection. At 3 weeks of age, the study pigs were vaccinated, in accordance with routine veterinary practice in Austria, against Mycoplasma hyopneumoniae (MycoFLEX; Boehringer Ingelheim, Ingelheim, Germany) and porcine circovirus type 2 (CircoFLEX; Boehringer Ingelheim). The animal infection experiments took place under biosafety level 2 containment at the University of Veterinary Medicine, Vienna, Austria. Transfer of animals and random allocation to infection and control groups was done 1 week prior to FLUAVsw/PBS application (day Ϫ6  (Fig. 1). One infected pig and one control pig were euthanized at day 2 postinfection (p.i.). Five infected pigs and five control pigs were euthanized consecutively on days 4, 6, 9, 12, 15, and 42 to 46 p.i. (days 42 to 46 p.i. are referred to here as "day 44") ( Fig. 1). For euthanasia, pigs were anesthetized (Narketan and Stresnil, both i.m.), and T61 (embutramide, mebezonium iodide, and tetracaine hydrochloride at 1 ml/10 kg BW; MSD, Whitehouse Station, NJ) was administered intracardially. Swine leukocyte antigen (SLA) haplotyping of pigs was performed as previously described (43). The low-resolution SLA haplotypes of all infected pigs are given in Table 1  split into three separate animal infection experiments, which were performed sequentially, to allow for efficient working practices. In each experiment, the animals were given an adaptation phase of 6 days, before FLUAVsw (infected, n ϭ 31) or PBS (control, n ϭ 31) was applied intratracheally. Application was performed under visual control of a bronchoscope. The photograph shows the tracheal bronchus (TB) and tracheal bifurcation (BF) as seen through the bronchoscope when positioned for application. For tissue collection, animals were euthanized on the indicated days postinfection. Blood samples were taken from both euthanized and surviving animals. The number of euthanized animals (infected plus control) is denoted separately for each study day and makes a total of 5 infected and 5 control animals per time point (with the exception of day 2 p.i., when only two pigs were euthanized). *, the late time points (days 42, 44, and 46) served to address memory responses of T cells; results obtained on these study days were combined and are referred to as "day 44 p.i." throughout this article.  Table 2). These isolates were obtained from nasal swabs of pigs with overt respiratory disease and were passaged twice in MDCK cells as previously described (44). Virus titers were determined in the same cell type using trypsin-supplemented, serum-free medium. Titer calculation with the Kaerber formula was based on cytopathic effects and staining of viral nucleocapsid antigen using the monoclonal antibody ATCC Hb-65. Full-length genome sequences were available from public databases or have previously been established by Sanger sequencing as described by Harder et al. (44). Sequence accession numbers are shown in Table 2. Lineage identity (avian versus human origin) was assigned according to the results of extensive homology searches using the Influenza Research Database (www.fludb.org). Pairwise comparison of identity measures of deduced protein sequences from all eight genome segments were obtained using GeneDoc software (www.nrbsc.org/gfx /genedoc/gdfeedb.htm).
Clinical and pathological examination. Clinical examinations, including the measurement of rectal temperatures, were carried out on a daily basis. A scoring system was used to document respiratory signs, behavior, and body condition ( Fig. 2A). At necropsy, gross lung lesions were documented, and samples from all seven lung lobes were fixed in 10% neutral buffered formalin for histological investigation. Paraffin sections (3 m) were stained with hematoxylin and eosin, and 11 different pathological parameters were quantified using scores from 0 (absent) to 3 (high grade) (Fig. 3A).
Sample collection and isolation of lymphocytes. Lung tissue, tracheobronchial lymph nodes (TBLN), mesenteric lymph nodes (MesLN), and bronchoalveolar lavage fluid (BALF) were collected at euthanasia. Blood samples were taken from the jugular veins of euthanized as well as surviving animals. Serum was frozen at minus 20°C for use in serum neutralization assays. Peripheral blood mononuclear cells (PBMC) were isolated from heparinized blood by density gradient centrifugation (Pancoll human; density, 1.077 g/ml; PAN Biotech) as described elsewhere (45). Tracheobronchial lymph nodes and mesenteric lymph nodes were processed for isolation of lymphocytes as previously described (46). For the isolation of lymphocytes from lung tissue, a sample of tissue (approximately 4 by 3 by 2 cm) from the center of the right cranial lobe was snipped into small pieces (approximately 3 by 3 by 3 mm), rinsed with PBS to reduce contamination from blood, and shake incubated at 37°C for 1 h in 25  ). Digested tissue was sieved through cotton wool, spun down (350 ϫ g, 10 min, room temperature), resuspended in 30 ml PBS, and layered onto 15 ml of lymphocyte separation medium (Pancoll human; PAN Biotech). After centrifugation (920 ϫ g, 30 min, room temperature) and collection of the interphase, cells were washed (350 ϫ g, 10 min, 4°C) twice with PBS and once with wash medium (RPMI 1640 with stable glutamine supplemented with 5% FCS, 100 IU/ml penicillin, and 0.1 mg/ml streptomycin). Cells seeded for in vitro restimulation were resuspended in cell culture medium (RPMI 1640 with stable glutamine supplemented with 10% FCS, 100 IU/ml penicillin, and 0.1 mg/ml streptomycin). Bronchoalveolar lavage fluid was obtained postmortem. An incision was made into the trachea, a tracheal tube was inserted, and the lung was flushed with 100 ml of PBS. BALF was spun down (350 ϫ g, 10 min, 4°C), and supernatant was frozen at minus 80°C for detection of viral RNA in subsequent quantitative reverse transcription-PCR (qRT-PCR) experiments.
Ex vivo determination of phenotype and Ki-67 expression. For ex vivo flow cytometric (FCM) analyses, freshly isolated cells from lung and tracheobronchial lymph nodes were suspended in PBS (without Ca 2ϩ and Mg 2ϩ ; PAN Biotech) supplemented with 10% (vol/vol) porcine plasma (in-house preparation) and adjusted to 1.5 ϫ 10 6 cells per sample. Staining was performed in 96-well round-bottom plates (with incubations for 20 min at 4°C). Antibodies and secondary reagents are listed in Table 3. A Live/Dead Near-IR stain kit (Invitrogen) was used for the discrimination of dead cells. For fixation and permeabilization, the Foxp3 staining buffer set (eBioscience, San Diego, CA) was used according to the manufacturer's instructions.
ELISpot assays for gamma interferon (IFN-␥) production. Enzymelinked immunosorbent spot (ELISpot) assays were performed as previously described (42). For in vitro restimulation, 3 ϫ 10 5 cells, freshly isolated from blood (PBMC), lung, tracheobronchial lymph nodes, and mesenteric lymph nodes, were incubated with FLUAVsw (infection strain; multiplicity of infection [MOI] ϭ 0.1) for 24 h at 37°C. Mesenteric lymph nodes (MesLN) were analyzed as control tissues. Medium and mock incubations served as additional negative controls, as did cells isolated from noninfected control animals. For restimulation experiments with heterologous virus and nucleoprotein, PBMC and lung cells were defrosted and incubated for 24 h with virus (MOI ϭ 0.3) or nucleoprotein (1 g/ml), or were mock infected, at 37°C.

Intracellular cytokine staining (ICS).
To analyze production of IFN-␥, tumor necrosis factor alpha (TNF-␣), and interleukin-2 (IL-2) by FLUAVsw-specific CD4 ϩ and CD8 ϩ T cells, freshly isolated cells from blood, lung, and tracheobronchial lymph nodes were restimulated in vitro with FLUAVsw (infection strain, MOI ϭ 0.1) or mock treated for 18 h at 37°C. For restimulation experiments including heterologous viruses and  Table 3.  from PBS-treated control pigs and FLUAVsw-infected pigs were stained with hematoxylin and eosin and examined for the presence and quantity of parameters A to K (defined in the figure). The quantity and presence of each parameter were recorded as a score (0 [absent] to 3 [high grade]). Dot sizes represent the score for one individual animal (day 2) or the mean score for five animals (days 4 to 44). (B) Histopathology scores (HPS) were summed for each individual animal (infected, diamonds; control, gray dots). Horizontal bars indicate the median value for five animals.
Single-stain samples were included for automated calculation of compensation. From the ex vivo stainings, 2 ϫ 10 5 lymphocytes were recorded using the high-throughput sampler of the flow cytometer. Cells subjected to overnight restimulation and intracellular cytokine staining were analyzed in FCM tubes, yielding between 5 ϫ 10 5 and 1 ϫ 10 6 recorded lymphocytes per sample. Doublet discrimination and dead-cell exclusion were performed as previously described (42). Data were processed by FACSDiva software (version 6.1.3; BD Biosciences) and transferred to Microsoft Excel (Office 2010; Microsoft, Redmond, WA) for further calculations.
Serum neutralization test. Infection strain neutralizing antibody titers in defrosted sera were analyzed as described elsewhere (47).
qRT-PCR. Viral loads in BALF were assessed following a TaqMan one-step quantitative reverse transcription-PCR (qRT-PCR) assay in a

RESULTS
Clinical signs, histopathology, and viral load. All 31 FLUAVswinoculated pigs developed clinical signs indicative of lower respiratory tract infection, with the most prominent signs being dyspnea, coughing, and abnormal breathing sounds upon lung auscultation ( Fig. 2B and data not shown). These respiratory signs were accompanied by pyrexia (rectal temperature of Ͼ40°C [physiological range for growing pigs, 38.7 to 39.3°C]), and lethargy. Inflammatory signs of the upper respiratory tract (nasal/ ocular discharge and conjunctivitis) were absent in almost all animals (data not shown). Mean summed clinical scores (CS) were highest on day 1 p.i., and within 1 week they gradually declined to levels similar to those of the PBS-treated control animals (Fig. 2C). Rectal temperatures showed a biphasic pattern, with an initial peak on day 1 p.i. followed by a second peak on day 3 p.i. (Fig. 2D). At necropsy, gross lesions were restricted to the lungs of infected animals (1 to 10% of total lung surface). During the acute phase of infection (day 4 to day 15 p.i.), gross lung lesions included tissue consolidation, atelectasis, and emphysema. At 6 weeks p.i. (days 42, 44, and 46), fibrotic alterations were observed in the lungs of four out of five animals (data not shown). These findings were confirmed microscopically, where a scoring system was used to quantify 11 different alterations. An overview of the 11 different histological alterations, their quantity, and their location with respect to the lung lobes is given in Fig. 3A. The most prominent findings were, among others, necrosis of bronchiolar epithelium (days 2 and 4 p.i.), obstruction of bronchiolar lumina, alveolar histiocytosis, inflammatory cells in alveolar walls, and atelectasis. Intralobular fibrosis, an indicator of tissue repair, was observed from day 6 to 9 p.i. onwards. Generally, accessory, cranial, and middle lung lobes were affected more severely than caudal lobes. By the end of the study, on day 44 p.i., lungs of infected animals showed residual hyperplasia of bronchus-associated lymphoid tissue (BALT), atelectasis, and intralobular fibrosis but had otherwise completely recovered to a level comparable to that of the PBS-treated control animals. Summed histopathology scores of individual pigs are shown in Fig. 3B.
The viral load, as measured by qRT-PCR from BALF of euthanized animals, decreased markedly from day 4 to day 6 p.i. (Fig.  4A). In the BALF of animals euthanized from day 9 p.i. onwards, viral RNA was below the detection limit. Control animals remained negative in qRT-PCR throughout the experiment.
Neutralizing antibodies in serum. In sera from seven out of 30 infected animals, neutralizing antibodies against the infection strain could be detected as early as 4 days p.i. (Fig. 4B). These early responders also developed the highest antibody titers on consecutive days. By day 6 p.i., all but one of the remaining animals (24 out of 25) had detectable levels of neutralizing antibodies. In most animals, neutralizing antibodies reached their highest levels between days 9 and 15 p.i. Notably, 4 weeks later (day 44 p.i.), antibody titers were only slightly reduced in the five remaining animals. No FLUAVsw-neutralizing antibodies were detected in the sera of PBS-treated control animals (data not shown).
Ex vivo Ki-67 expression and phenotype of CD4 ؉ and CD8 ؉ T cells. Expression analysis of the proliferation-associated molecule Ki-67 is widely used to track and quantify the expansion phase of T cells after an experimental vaccination or infection (25,(49)(50)(51). Having previously observed Ki-67 expression kinetics reminiscent of CD8 T-cell expansion in the blood of influenza virus-infected pigs (42), we now aimed to detect an increase of recently activated T cells in lung and lung-draining lymph nodes. For this purpose, we stained freshly isolated lymphocytes for expression of Ki-67 in combination with other activation/differentiation-related molecules (CD4 ϩ T cells, CD8␣ and CD27; CD8 ϩ T cells, perforin and CD27) and analyzed them by flow cytometry (Fig. 5A and D). In pigs, the alpha chain of the CD8 molecule is expressed on activated and antigenexperienced CD4 ϩ T cells (52). Porcine CD8 ϩ T cells, however, express higher levels of CD8␣ and can be identified by a CD4 Ϫ CD8␣ high CD8␤ ϩ/dim phenotype (53)(54)(55). For ex vivo analyses of CD8 ϩ T cells, CD8␣ high cells were gated (first contour plot of Fig.  5D), as weak performance of the CD8␤-Zenon-Alexa 488 staining was experienced in combination with the Foxp3 staining buffer, particularly with respect to lung lymphocytes.
For CD4 ϩ T cells, an increase of Ki-67 ϩ CD8␣ ϩ CD27 ϩ cells was most pronounced in the TBLN from day 4 to 9 p.i., with only a slight increase at day 6 p.i. in the lung (Fig. 5B). In contrast, for CD8 ϩ T cells, the increase in cells with an early effector phenotype (Ki-67 ϩ perforin ϩ CD27 ϩ ) was most evident in the lung from day 4 to day 12 p.i., with a clear peak on day 6 p.i. (Fig. 5E). Percentages of the respective subsets lacking CD27 expression showed high pig-to-pig variation and did not reveal such a clear time-dependent pattern (Fig. 5C and F).

IFN-␥ ELISpot assay following homologous viral restimulation in vitro.
Ex vivo analyses clearly pointed toward an accumulation of Ki-67 ϩ T cells with an early effector phenotype at the site of infection (Fig. 5). In order to screen for a FLUAVsw-specific response within PBMC, lungs, and TBLN, we performed IFN-␥ ELISpot assays upon in vitro restimulation (infection strain, MOI ϭ 0.1). The first IFN-␥ responses were detected within PBMC isolated on day 4 p.i. (Fig. 6). The median frequency of IFN-␥-producing PBMC peaked on day 9 p.i. and by day 15 returned to a level similar to that on day 4 p.i., where it remained until the end of the study on day 44 p.i. Within lung tissue, we observed a strong increase of IFN-␥-producing cells from day 4 to day 6 p.i. This elevated level of response in the lung more or less plateaued up to day 15 p.i. and was found to have fallen only slightly by day 44 p.i. In TBLN, FLUAVsw-specific IFN-␥ responses were first detected on day 6 p.i. and were still clearly detectable on day 44 p.i. No FLUAVsw-specific IFN-␥ production could be detected within cells isolated from the MesLN or within any cells isolated from PBS-treated control animals ( Fig. 6 and data not shown).
FLUAVsw-specific production of IFN-␥, TNF-␣, and IL-2 by CD4 ؉ and CD8 ؉ T cells. IFN-␥ ELISpot assays revealed reactivity to viral restimulation in PBMC, lungs, and TBLN of FLUAVswinfected animals. However, it should be noted that other cells (besides T cells) capable of IFN-␥ production, such as NK cells, may have contributed to the IFN-␥ production measured by ELISpot assays. To specifically address the phenotype and multifunctionality of responding T cells, we performed intracellular cytokine staining (ICS) for the expression of IFN-␥, TNF-␣, and IL-2 following in vitro restimulation (infection strain, MOI ϭ 0.1). Figures 7 and 8 show representative cytokine expression in CD4 ϩ T cells and CD8 ϩ T cells, respectively. Boolean gating revealed that FLUAVsw-specific IFN-␥-positive CD4 ϩ T cells appeared in blood and lung as early as 4 days p.i. (Fig. 9). From day 6 or 9 p.i. onwards, singly positive TNF-␣ ϩ and doubly positive IFN-␥ ϩ TNF-␣ ϩ CD4 ϩ T cells could also be detected at all sites, though frequencies differed. Triply positive IFN-␥ ϩ TNF-␣ ϩ IL-2 ϩ CD4 ϩ T cells could be determined in TBLN and PBMC already at day 6 p.i. but did not appear in the lung before day 9 p.i. The lung was clearly dominated by singly positive IFN-␥ ϩ cells, with doubly positive IFN-␥ ϩ TNF-␣ ϩ cells being the second most abundant subset. Singly positive TNF-␣ ϩ CD4 ϩ T cells markedly  accumulated in the TBLN. The highest frequencies of triple-cytokine-producing CD4 ϩ T cells were found within PBMC, followed by TBLN, with peak frequencies detected at 9 to 12 days p.i. At the latest assessment point, i.e., 44 days p.i., FLUAVsw-specific responses comprised primarily IFN-␥ single producers in the lung, TNF-␣ single producers in the TBLN, and IFN-␥/TNF-␣/IL-2 triple producers in blood.
Similar to the case for CD4 ϩ T cells, the first detectable FLUAVsw-specific CD8 ϩ T cells were IFN-␥ single producers at day 4 p.i.; however, at this early time point, this cell subset was limited to the lung (Fig. 10). For these singly positive IFN-␥ ϩ CD8 ϩ T cells from the lung, a highly increased response could be detected from day 6 p.i. onwards, leading to frequencies that were up to 30 times as high as those found in PBMC or TBLN. The frequency of IFN-␥/TNF-␣ double producers was also highest in the lung. At 44 days p.i., around 1% and 0.5% of CD8 ϩ T cells in the lung continued to respond with IFN-␥ production and IFN-␥/TNF-␣ coproduction, respectively. In contrast to CD4 ϩ T cells, CD8 ϩ T cells failed to produce substantial amounts of IL-2 (Fig.  8). Both CD4 ϩ and CD8 ϩ T cells isolated from the mesenteric lymph nodes remained unresponsive to in vitro restimulation with FLUAVsw, as assessed by the lack of intracellular cytokine accumulation (Fig. 7 to 10). To confirm the influenza virus specificity of cytokine production in CD4 ϩ and CD8 ϩ T cells, we com-pared the in vitro response to FLUAVsw with the response to an unrelated live attenuated PRRSV strain, which is, like influenza virus, an enveloped RNA virus. Defrosted PBMC initially isolated at days 4 and 12 p.i. were restimulated in parallel with the FLUAVsw infection strain (MOI ϭ 0.1) and the PRRSV strain (MOI ϭ 0.2). The frequencies of IFN-␥-, TNF-␣-, and/or IL-2producing CD4 ϩ and CD8 ϩ T cells in PRRSV-stimulated microcultures never exceeded those frequencies found in influenza- mock-stimulated cultures (data not shown), indicating that the mock-corrected frequencies of cytokine-producing CD4 ϩ and CD8 ϩ T cells in Fig. 9 and 10 represent those of influenza virusspecific T cells. Expression of CD27 on cytokine-defined subsets of CD4 ؉ and CD8 ؉ T cells. As CD27 has been proposed to discriminate between differentiation-related stages of porcine CD4 ϩ and CD8 ϩ T cells (56,57), we looked for differences in the expression of CD27 at different time points postinfection and between different cytokine-producing T-cell subsets. As shown in the top row of Fig. 11A, total CD4 ϩ T cells within PBMC and TBLN were primarily CD27 ϩ , whereas total CD4 ϩ T cells isolated from the lung were largely negative for CD27. When looking at cytokine-producing subsets of CD4 ϩ T cells, we found that in PBMC and lungs, the ratio of CD27 ϩ to CD27 Ϫ cells was around 1:1, irrespective of the cytokines produced. For the lung this indicates that, from day 6 p.i. onwards, cytokines were preferentially produced by CD27 ϩ cells.
This observation was even more pronounced for CD8 ϩ T cells in the lung (Fig. 11B), where total CD8 ϩ cells had a balanced 1:1 expression of CD27, and singly positive IFN-␥ ϩ and doubly positive IFN-␥ ϩ TNF-␣ ϩ cells were clearly enriched in the CD27 ϩ subset. Therefore, overall FLUAVsw-specific cytokine production was dominated by T cells with an early effector phenotype or central memory phenotype across all analyzed time points.
Cross-reactivity of CD4 ؉ and CD8 ؉ T cells. In order to test for cross-reactivity of FLUAVsw-specific T cells, defrosted PBMC and lung cells from seven infected animals, isolated from 9 days p.i. to 44 days p.i., were restimulated with heterologous FLUAVsw isolates or recombinant influenza A virus nucleoprotein (rNP), and cytokine responses were assessed by IFN-␥ ELISpot assay and ICS for IFN-␥, TNF-␣, and IL-2 (Fig. 12).
As shown in Table 2, the internal segments of the FLUAVsw isolate used for infection and of the three FLUAVsw isolates used for heterologous in vitro restimulation all belonged to the avianderived European lineage of FLUAVsw. Analysis of HA and NA sequences assigned the isolates to three different HA and two NA subtypes/lineages. Deduced protein sequences revealed a very high degree of identity for the internal proteins of the FLUAVsw isolates (97% or higher) except for the NS-1 ( Table 4). The recombinant NP derived from an H7N9 isolate had a homology of at least 94% with the NPs of all other strains.
All heterologous virus strains (Fig. 12C to E) as well as rNP (Fig. 12F) were able to evoke cytokine production in restimulated cells. The frequency of responding cells was lower than with homologous restimulation (Fig. 12B) and was generally higher in the lung than in PBMC. When looking at the frequency of IFN-␥ producers, CD4 ϩ T cells were present at somewhat higher frequencies than CD8 ϩ T cells, but the latter tended to contain a higher proportion of IFN-␥-/TNF-␣-coproducing cells in the lung. At 44 days p.i., CD8 ϩ T cells from the lung also demonstrated higher heterologous responses than their CD4 ϩ counterparts ( Fig. 12A and E [squares]). Cross-reactive IFN-␥-/TNF-␣-/ IL-2-triple-producing T cells could be identified only at low frequencies within blood-derived CD4 ϩ T cells for H1avN1 and H1avN2 ( Fig. 12C and D), but not for H3N2 (Fig. 12E), and were absent within CD8 ϩ T cells. The latter finding is consistent with the homologous restimulations presented in Fig. 8.
The CD4 response to rNP was very similar to the response following viral restimulation. Notably, an rNP-specific IFN-␥ response was also detected in CD8 ϩ T cells isolated from the lungs in 4 out of 5 infected animals, although delivery of rNP-derived peptides to the major histocompatibility complex (MHC) class I presentation pathway may have been suboptimal. In summary, these findings indicate the presence of cross-reactive CD4 ϩ and CD8 ϩ T cells at various time points postinfection, including day 44, in particular within lung tissue but also, at low frequencies, within PBMC.

DISCUSSION
In three controlled animal infection experiments, we analyzed the location, quality, and cross-reactivity of porcine influenza virusresponsive T cells. We were able to demonstrate that both CD4 ϩ and CD8 ϩ influenza virus-reactive T cells accumulate in the lungs of infected pigs from day 4 p.i. onwards, produce mainly IFN-␥ and/or TNF-␣, and form cross-reactive memory at the site of infection.
Following intratracheal application of virus, all 31 infected pigs developed fever and showed clinical signs of lower respiratory tract infection. Lung pathology revealed necrosis of bronchiolar epithelia until day 6 p.i., which is consistent with productive replication of influenza virus in these cells. Viral RNA in BALF fell from day 4 to day 6 p.i. and was undetectable at the remaining sampling points. Clearance of influenza virus by day 7 p.i. has been reported for influenza virus-infected swine (58), and this is consistent with the observed rapid recovery of influenza virusinfected pigs. Both neutralizing antibodies in blood and influenza virus-specific T cells in lung tissue could already be detected at low titers/frequencies at 4 days p.i. and showed a substantial increase by day 6 p.i., indicating their participation in the early control of viral infection. This early activation of T cells in a primary immune response is surprising, but NP-specific CD8 ϩ T cells have been detected previously by tetramer staining or IFN-␥ production in mediastinal lymph nodes and lungs of influenza virusinfected mice as early as 3 to 5 days p.i. (59,60). Similarly, for lymphocytic choriomeningitis virus (LCMV)-infected mice, tetramer staining demonstrated the presence of LCMV NP-specific CD8 ϩ T cells in the spleen from day 3 p.i. onwards (61).
Both CD4 ϩ and CD8 ϩ T cells isolated postinfection showed robust (multi-)cytokine responses to in vitro restimulation with homologous/heterologous virus and influenza virus nucleoprotein. CD8 ϩ T cells produced mainly IFN-␥ and/or TNF-␣ and lacked significant IL-2 production, confirming previous observations with blood-derived CD8 ϩ T cells from influenza virus-infected pigs (42) and nonhuman primates (62). In humans, preexisting IFN-␥ ϩ CD8 ϩ T cells lacking IL-2 production, and not IFN-␥ ϩ IL-2 ϩ CD8 ϩ T cells, could be correlated with decreased symptom scores upon natural infection with pandemic H1N1 influenza virus (26), arguing against the necessity of CD8-derived IL-2 for clinical protection. Indeed, IFN-␥ is often described as a key antiviral cytokine. Mouse studies have demonstrated the importance of rapid IFN-␥ production by airway-resident memory CD8 ϩ T cells for protection against heterologous influenza virus challenge (22). In addition, CD4-derived IFN-␥ was shown to be necessary for the generation of lung-resident CD8 T-cell memory (63). Thus, the dominant populations of FLUAVsw-specific IFN-␥-producing CD4 ϩ and CD8 ϩ T cells identified in various organs in this study are likely to have contributed to controlling the infection.
Multifunctionality, i.e., the simultaneous production of more than one cytokine or effector molecule at the single-cell level, has been proposed as an important quality measure of T-cell responses (64,65). Perforin is one of the main effector molecules of CD8 ϩ T cells, mediating apoptosis of influenza virus-infected cells and viral clearance (16). Evidence that multifunctionality of CD8 ϩ T cells may translate into increased cytotoxic activity has been provided by Lichterfeld et al. (66), who found increased perforin expression and stronger ex vivo cytotoxic activity of CD8 ϩ T cells producing both IFN-␥ and TNF-␣ than CD8 ϩ T cells producing IFN-␥ only. In the present study, we were able to detect a considerable proportion of doubly IFN-␥/TNF-␣-producing CD4 ϩ and CD8 ϩ T cells in porcine lungs infected with FLUAVsw. Moreover, we could detect peak frequencies of proliferating (Ki-67 ϩ ) and perforin-expressing CD8 ϩ T cells in lungs at 6 days p.i., clearly suggesting that CD8 ϩ T cells play a role in viral clearance by cytolytic activity.
In order to distinguish between differentiation stages of influenza virus-responsive T cells, we coanalyzed CD27 expression. Porcine CD27 was shown to be differentially expressed on antigen-experienced (CD8␣ ϩ ) CD4 ϩ T cells, and functional analyses suggested a division into central memory (CD27 ϩ ) and effector memory (CD27 Ϫ ) cells (56). For perforin-expressing CD8 ϩ T cells in pigs, age-related changes suggested a division into early (CD27 ϩ ) and late (CD27 Ϫ ) effector cells (57). In mice, engagement of costimulatory CD27 has been shown to promote effector expansion as well as survival and memory transition of T cells (67)(68)(69). Ballesteros-Tato et al. (59) correlated CD27 expression with the proliferation of effector CD8 T cells by showing that the in vivo expansion of influenza virus nucleoprotein-specific CD8 ϩ T cells was dependent on CD70-CD27 interaction. Our ex vivo observations support this finding and extend it to include CD4 ϩ T cells, as kinetics reminiscent of effector expansion were evident only in Ki-67 ϩ T cells expressing CD27. We also addressed the expression of CD27 on influenza virus-specific T cells over the course of the infection and with respect to the cytokines produced.
In the lung, we found that cytokine-producing influenza virusspecific T cells were clearly skewed toward CD27 expression. This propensity for CD27 expression was most pronounced for IFN-␥producing CD8 ϩ T cells and may be related to better survival and memory transition, supported by the high frequencies of IFN-␥producing CD8 ϩ T cells both in the acute phase and 5 weeks after clearance of infection. Indeed, the high numbers of FLUAVsw-specific T cells in lungs, long after clearance of infection, strongly suggest the formation of lung-resident T-cell memory, which has been shown to mediate superior protection against influenza virus challenge in mice (19)(20)(21)(22). We also detected considerable memory responses in lung-draining lymph nodes and peripheral blood, possibly indicating immunological memory that can be rapidly recruited to the lung to support frontline memory T cells upon secondary infection (70).
In contrast to most neutralizing antibodies, T cells have the potential to be broadly reactive across influenza virus subtypes (14). T cells recognize internal epitopes that, due to functional constraints (71), are less likely to accumulate nucleotide substitutions and are therefore highly conserved among different influenza virus strains. Although cross-reactive T-cell memory cannot confer sterilizing immunity, their presence has been shown to be associated with reduced disease severity in humans (25)(26)(27). For pigs, reduced viral replication and shedding have previously been shown in consecutive experimental infections with different influenza A virus strains (36,(72)(73)(74)(75). One study demonstrated that this was associated with an increase of CD8 ϩ T cells in lung tissue (33). Moreover, by infecting pigs with a genetically modified influenza virus strain that suppresses the HA signal sequence (S-FLU) (76) and that does not induce neutralizing antibodies, Morgan et al. (41) found a correlation between IFN-␥-producing CD8 ϩ and CD4 ϩ T cells in BALF and a reduction of viral load in nasal swabs and lungs upon homologous and heterologous challenge. In the present study, we demonstrated T-cell reactivity against three heterologous influenza virus strains and nucleoprotein. Influenza virus nucleoprotein evoked CD4 T-cell responses comparable to those of heterologous virus, in both magnitude and quality, suggesting an immunodominant role of nucleoprotein for porcine CD4 ϩ T cells.
Overall, our data support the idea that in the pig, a natural influenza virus host, T cells are actively involved both in the clearance of primary infection and in the control of heterologous reinfection. We are confident that this and ensuing work on the porcine T-cell response to influenza virus will strengthen the position of pigs as large-animal models for human influenza virus infection and will immediately aid the development of more effective vaccines for swine. Improved control of influenza virus replication in pigs will reduce the chances of influenza virus reassortments, as well as reducing pig-human transmission, thereby lowering the risk of swine-mediated pandemic outbreaks.