Human infections with highly pathogenic avian influenza A (H5N1) virus are frequently fatal but the mechanisms of disease remain ill-defined. H5N1 infection is associated with intense production of proinflammatory cytokines, but whether this cytokine storm is the main cause of fatality or is a consequence of extensive virus replication that itself drives disease remains controversial. Conventional intratracheal inoculation of a liquid suspension of H5N1 influenza virus in nonhuman primates likely results in efficient clearance of virus within the upper respiratory tract and rarely produces severe disease. We reasoned that small particle aerosols of virus would penetrate the lower respiratory tract and blanket alveoli where target cells reside. We show that inhalation of aerosolized H5N1 influenza virus in cynomolgus macaques results in fulminant pneumonia that rapidly progresses to acute respiratory distress syndrome with a fatal outcome reminiscent of human disease. Molecular imaging revealed intense lung inflammation coincident with massive increases in proinflammatory proteins and IFN-α in distal airways. Aerosolized H5N1 exposure decimated alveolar macrophages, which were widely infected and caused marked influx of interstitial macrophages and neutrophils. Extensive infection of alveolar epithelial cells caused apoptosis and leakage of albumin into airways, reflecting loss of epithelial barrier function. These data establish inhalation of aerosolized virus as a critical source of exposure for fatal human infection and reveal that direct viral effects in alveoli mediate H5N1 disease. This new nonhuman primate model will advance vaccine and therapeutic approaches to prevent and treat human disease caused by highly pathogenic avian influenza viruses.
Human infection with highly pathogenic avian influenza A viruses of the H5N1 subtype is characterized by severe pneumonia that frequently progresses to acute respiratory distress syndrome and death (1). Since its re-emergence in 2003 there have been >850 confirmed cases of human H5N1 infection worldwide with a cumulative case fatality rate of >50%. The virus is an emerging global health threat, with 173 confirmed human cases in Egypt in 2014–2015 associated with widespread H5N1 infection in poultry (http://www.who.int/influenza/human_animal_interface/en/). Highly pathogenic avian H5 viruses were detected in the United States for the first time in 2014 and caused widespread outbreaks and death in poultry (2), although these viruses currently present a low risk for animal-to-human transmission (3).
The pathogenesis of human H5N1 pneumonia and the relative contribution of virus replication and immunopathology in driving disease remain unclear, and this has fundamental implications in how we treat severe H5N1 infections. H5N1 virus infects human alveolar epithelial cells (AEC) (4–9), and loss of alveolar barrier function through AEC death (6, 10) could be a central factor in progressive H5N1 pneumonia. H5N1 also binds to and infects alveolar macrophages in the human lower respiratory tract (4, 8). Alveolar macrophages play a critical role in protecting the lung from damage (11), but the impact of highly pathogenic H5N1 infection on alveolar macrophage survival in vivo has not been evaluated. Markedly increased plasma levels of proinflammatory cytokines and chemokines, including IP-10 (CXCL10), MCP-1 (CCL2), and IL-6, are a hallmark of human H5N1 infection (12–15). H5N1 induces more profound cytokine responses from AEC and macrophages in vitro than do less pathogenic isolates (13, 16–18), supporting the hypothesis that a harmful cytokine storm is central to disease pathogenesis (17, 19). However, whether proinflammatory cytokines drive disease or are a consequence of virus replication that itself damages the lung remains controversial (20).
Studies in nonhuman primates have substantially advanced our understanding of human infection with highly pathogenic H5N1 viruses largely through genome- and transcriptome-based analyses but have not addressed the cellular factors mediating disease (21–26). A significant issue with nonhuman primate studies to date is that unlike human disease, infection causes death in only a small minority of animals (21–23, 26). All published nonhuman primate studies have used an inoculation method involving application of virus in liquid suspension (21–26) or via large droplets (27) into the trachea, often in combination with other mucosal routes. These delivery methods are unlikely to expose target cells in the alveoli to significant quantities of virus because liquid suspensions and/or large droplets are cleared by the mucociliary apparatus of the upper respiratory tract. In contrast, small-particle aerosols of virus of ≤5 μm in diameter remain in suspension in air for sustained periods of time and penetrate to the lower respiratory tract when inhaled (28, 29). Studies in the ferret model show that aerosol inoculation resembles a natural, airborne infection with H5N1 virus that is highly infectious and lethal (30, 31); however, this route of exposure has not been attempted in nonhuman primates.
In this study we explored the use of small-particle aerosols in H5N1 infection of macaques. We report that infection with aerosolized influenza A/Vietnam/1203/2004 (H5N1) virus causes fulminant pneumonia that rapidly progresses to the acute respiratory distress syndrome, recapitulating fatal human disease. Our data indicate that direct damage to the alveolus is the central factor in disease pathogenesis, as widespread infection of cells in the alveoli leads to massive alveolar macrophage depletion and AEC apoptosis that significantly compromises the crucial barrier function of the alveolus. Our findings indicate that aerosolized H5N1 virus blankets target cells in alveoli and is likely a major and unappreciated source of exposure in fatal human infections. The work establishes a new nonhuman primate disease model for evaluation of vaccine and therapeutic approaches to prevent and treat infection with highly pathogenic avian influenza viruses.
Materials and Methods
Study approval and oversight
Experiments were performed in the University of Pittsburgh Regional Biocontainment Laboratory BSL-3 facility, which is a registered entity with the Centers for Disease Control and Prevention/U.S. Department of Agriculture for work with highly pathogenic avian influenza viruses. The study was conducted in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All experiments had oversight and approval by the Institutional Animal Care and Use Committee (protocol no. 15055917) and the Institutional Biosafety Committee (protocol no. 055-15) at the University of Pittsburgh. All personnel wore personal protective equipment and used powered air-purifying respirators (3M Versaflo) or worked in a class III biological safety cabinet.
Seven healthy adult female cynomolgus macaques (Macaca fascicularis) were confirmed to be seronegative for influenza A nucleoprotein Abs by ELISA (Virusys) and were surgically implanted with a TA10TA-D70 telemetry device (Data Sciences International, St. Paul, MN) placed in the abdominal wall to monitor temperature and activity as described (32). Three additional cynomolgus macaques were used as naive controls.
H5N1 virus inoculum
Highly pathogenic avian influenza A/Vietnam/1203/2004 (H5N1) virus was provided by Dr. S. Mark Tompkins at the Department of Infectious Diseases, University of Georgia. Virus propagation was carried out under biosafety level 3 conditions at the University of Georgia. Virus stock was amplified in 10- to 11-d-old embryonated specific pathogen-free chicken eggs (Charles River Laboratories) for 24 h as previously described (33). Allantoic fluid from inoculated eggs was clarified by centrifugation and aliquots were stored at −80°C until transfer to the University of Pittsburgh Regional Biocontainment Laboratory biosafety level 3 facility. Virus titer was determined in Madin–Darby canine kidney (MDCK) cells by plaque assay and expressed as PFU per milliliter. Whole-genome sequencing of virus seed and propagated stocks was performed using the Illumina MiSeq next-generation sequencing platform. A comparison of the consensus sequence to publicly available A/Vietnam/1203/2004 (H5N1) sequences on the Influenza Research Database (http://www.fludb.org/) showed no amino acid changes in the virus seed and propagated stocks used in this experiment. Virus was diluted in media containing BSA, HEPES buffer, penicillin/streptomycin, and trypsin immediately prior to aerosol exposure.
Macaques were sedated with ketamine and respiratory function, including calculation of minute volume, was assessed by plethysmography, using a head-out chamber controlled by Buxco XA software (Data Sciences International). Macaques were then placed inside a class III biological safety cabinet (Baker) and sedation was maintained by constant low-dose infusion of ketamine. The head was inserted into a head-only exposure chamber (Biaera Technologies, Hagerstown, MD). Aerosols of virus were created using an Aeroneb Solo nebulizer (Aerogen, Deerfield, IL) and all aspects of exposure were controlled by the AeroMP exposure system (Biaera Technologies). Exposure duration ranged from 15 to 34 min (mean time, 24.3 min) and was determined by each macaque’s minute volume, the desired presented dose, the nebulizer concentration of the virus, and the spray factor of the virus as determined by sham exposures using virus without animals as previously described (34). Aerosol samples were collected in an all-glass impinger (Ace Glass, Vineland, NJ) containing viral growth media and Antifoam A (Sigma-Aldrich, St. Louis, MO) and analyzed by plaque assay to determine the actual aerosol concentration administered. Particle size was measured using an Aerodynamic Particle Sizer (TSI, Shoreview, MN) and confirmed to be an average of 4 μm in diameter.
Molecular imaging and animal sampling
Molecular imaging with 2-deoxy-2-[18F]fluoro-d-glucose (FDG) positron emission tomography (PET) and computed tomography (CT) using a microPE Focus 220 preclinical PET scanner (Siemens Molecular Solutions, Knoxville, TN) and an eight-slice helical CT scanner (Neurologica, Danvers, MA) was done as described (35) both before infection and at day 2 postinfection (PI) on all animals. Nasal saline washes and blood samples were collected preinfection and on days 1, 2, 4, and 6, depending on survival, and at the time of necropsy. Whole blood was analyzed for absolute cell counts and processed for plasma and PBMC as described previously (36). Bronchoalveolar lavages (BAL) were performed >2 wk preinfection and on days 2, 4, and 6, depending on survival. Whole BAL and BAL fluid samples were aliquoted and stored at −80°C. At the time of sacrifice, animals underwent transcardiac saline perfusion, and at necropsy various tissues were collected and either snap frozen in liquid nitrogen or fixed in 10% buffered formalin. Lung samples for detection of virus titers were taken from the extremities of upper, middle, and lower lobes and not from deep within the tissues. Alternatively, tissues were fixed in 2% paraformaldehyde for 4 h and infused with 30% sucrose overnight prior to cryopreservation as described (37).
Snap-frozen tissues were homogenized in media containing FBS using an Omni tissue homogenizer (Omni International). For RT-PCR, serum, BAL, or tissue homogenate was mixed with TRI Reagent (Ambion), and RNA was extracted using either a PureLink viral RNA/DNA extraction kit (Invitrogen) or RNeasy mini kit (Qiagen). The SuperScript III Platinum one-step quantitative RT-PCR kit (Invitrogen) was used for amplification of each RNA sample. Influenza A primers, probe, and cycling conditions used for real-time RT-PCR were followed as described in BEI Resources product information sheet NR-15592. A standard curve was generated using 10-fold dilutions of RNA from virus stock of known titer. For quantitation through plaque assay, nasal wash, BAL fluid, or homogenized tissue samples were adsorbed onto confluent monolayers of MDCK cells (American Type Culture Collection) in duplicate wells of a six-well plate and overlaid with agarose-containing media with TPCK-trypsin. MDCK cells were newly ordered, did not contain mycoplasma, and were used within five passages of thawing. Plates were incubated for 3 d at 37°C. Cells were fixed in formaldehyde, agar plugs were removed, and cells were stained with crystal violet to identify plaques.
Generation of lung suspensions
Lung cell suspensions were generated as described (38) with minor modifications. Briefly, intact lung was filled via the main bronchus with HBSS with Ca/Mg and HEPES buffer containing 12.9 U/ml elastase and 100 μg/ml DNase I without prior flushing, clamped closed, and incubated at 37°C. Lung tissue was manually disrupted and single cells were passed through a metal sieve and a 70-μm cell strainer. Cells were either freshly stained with Abs as described below or cryopreserved for later analysis.
Flow cytometric analyses
The following Abs were used to stain blood, BAL, and lung suspension and were from BD Biosciences unless noted otherwise: CD3 (clone SP34-2) and CD20 (2H7) (combined as lineage markers), CD4 (L200), CD11b (ICRF44), CD11c (S-HCL-3), CD14 (M5E2), CD16 (3G8), CD45 (DO58-1283), CD123 (7G3), CD163 (GHI/61), CD206 (19.2), active caspase-3 (C92-605), cytokeratin (CAM5.2), Ki-67 (B56), MHC class II (L243), and influenza A virus nucleoprotein (DPJY03; BEI Resources). The anti-nucleoprotein Ab was directly conjugated with the APEX Alexa Fluor 647 Ab labeling kit (Thermo Fisher). Flow cytometric analysis was performed as previously described (39). Absolute counts of blood mononuclear cells, CD4+ T cells, and intermediate monocytes were determined through a quantitative flow cytometry–based assay using BD Trucount tubes (40). Cell viability in BAL and lung suspensions was analyzed immediately after processing through trypan blue staining and determined to be >98% in each preparation. Between 1 and 2 × 106 cells were stained within 30 min of completing sample processing and were maintained on ice for the entirety of the stain. Matched isotype control Abs were used to define specific staining above autofluorescence. Where appropriate, cellular debris was excluded by use of forward scatter area and side scatter area gating. All samples were run on a BD FACSAria using BD FACSDiva software in the Regional Biocontainment Laboratory and analysis was performed using FlowJo software version 9.7.7 (Tree Star).
Analysis of cytokine and albumin concentrations
A 13-plex nonhuman primate-specific cytokine/chemokine LEGENDplex bead array (BioLegend) was used according to the manufacturer’s protocol on plasma and BAL fluid samples. All samples were run on a BD FACSAria, and analysis was performed using LEGENDplex analysis software (BioLegend). An ELISA specific for cynomolgus/rhesus IFN-α2 was performed using BAL samples according to the manufacturer’s protocol (VeriKine; PBL Assay Science). BAL fluids from postinfection time points were diluted 50-fold to remain within the standard limits of detection. An anti-monkey albumin ELISA was used to quantify albumin concentrations in the BAL fluid and plasma (Innovative Research). Plasma was diluted according to the manufacturer’s protocol and BAL fluid was diluted either 1000-fold for preinfection samples or 20,000-fold for postinfection samples.
In situ hybridization and immunofluorescence
In situ hybridization of formalin-fixed tissues was performed using 35S-labeled riboprobes synthesized using templates derived from 760 bp of influenza A/California/04/2009 matrix protein as previously described (41). For immunohistochemistry, sections were blocked in 5% serum prior to adding primary Ab for 1.5 h at 4°C. The following Abs were used: mouse anti-pan cytokeratin (AE1/AE3; Abcam), mouse anti-human IFN-α (MMHA-2; PBL Interferon Source), rabbit anti-human IFN-α2 (Thermo Scientific), mouse anti-human calprotectin Ab-1 (MAC 387; Thermo Scientific), mouse anti-human MCP-1 (5D3-F7; BioLegend), mouse anti-human MCP-1 biotin (5D3-F7, BioLegend), mouse anti-human CD163 biotin (eBioGHI/61; eBioscience), anti-influenza A nucleoprotein (DPJY03; BEI Resources), and mouse anti-human CD206 (15-2; AbD Serotec). Secondary Abs were from Thermo Fischer Scientific and included goat anti-mouse IgG1 Alexa Fluor 546, goat anti-mouse IgG2a Alexa Fluor 647, donkey anti-rabbit IgG Alexa Fluor 488, and donkey anti-rabbit IgG Alexa Fluor 546. Biotinylated Abs were detected using the HRP-streptavidin and Alexa Fluor 488 tyramide signal amplification kit. Cell nuclei were identified using membrane-permeable Hoescht dye, and slides were mounted in Gelvatol mounting media and viewed on an Olympus FluoView FV1000 confocal microscope. All staining included an appropriately concentrated isotype-matched control Ab.
All statistical analyses were performed in Prism v6.0 (GraphPad Software). A p value <0.05 was considered significant. All analyses were performed using two-tailed, nonparametric tests, either between control and infected animals or between paired samples. Significant p values from one-way ANOVA tests were followed up by a Dunn multiple comparison tests. A Grubbs test was used to determine whether any data points were considered to be significant outliers. The averages of log10 values were calculated as geometric means.
Aerosolized H5N1 infection of macaques causes fulminant pneumonia
We exposed seven adult cynomolgus macaques that were seronegative to influenza A viruses to small-particle aerosols of highly pathogenic avian influenza A/Vietnam/1203/2004 (H5N1), a well-defined strain isolated from a fatal human case that is commonly used in this macaque species (21–23, 26). Sedated macaques in a head-only chamber within a class III biological safety cabinet were exposed to aerosols averaging 4 μm in diameter that were generated using an electronic vibrating nebulizer. The mean challenge dose was 6.72 log10 PFU (Table I), which is similar to the amount given by intratracheal inoculation in previous reports using this strain (21–23, 26). All animals had a fever that peaked within 24 h of infection as determined by radiotelemetry (Fig. 1A). Infection had almost immediate effects on respiratory function as determined by plethysmography, with increases in respiration rate and decreases in tidal volume and expiratory time by 1–2 d PI, consistent with acute lower respiratory tract disease (Fig. 1B). To visualize the effects of infection on the lung, we performed molecular imaging, combining PET using the radiotracer FDG to label metabolically active cells with CT to define anatomy. Imaging at day 2 PI revealed severe, diffuse inflammation in the lungs of all animals, with significant increases in FDG uptake and lung volume relative to preinfection time points (Fig. 1C, 1D). These findings are consistent with the acute respiratory distress syndrome that characterizes progressive human H5N1 infection (1).
|.||.||.||Virus Titer (log10 PFU/ml or PFU/g)|
|Animal ID .||Virus Dose (log10 PFU) .||Time to Death (d) .||Day 1 .||Day 2 .||Day 4 .||Day 6 .||Day 2 .||Day 4 .||Day 6 .||Upper .||Middle .||Lower .||Upper .||Middle .||Lower .|
|.||.||.||Virus Titer (log10 PFU/ml or PFU/g)|
|Animal ID .||Virus Dose (log10 PFU) .||Time to Death (d) .||Day 1 .||Day 2 .||Day 4 .||Day 6 .||Day 2 .||Day 4 .||Day 6 .||Upper .||Middle .||Lower .||Upper .||Middle .||Lower .|
—, not available
Aerosolized infection is associated with systemic H5N1 virus replication
Respiratory disease rapidly progressed in all animals, with three animals dying and four animals being humanely sacrificed between 2 and 6 d PI (mean, 3.2 d) with respiratory failure (Table I). At necropsy, lungs were intensely hemorrhagic, exhibited intra-alveolar exudate and cellular consolidation, and were grossly enlarged, increasing 3- to 4-fold in weight relative to body weight–matched uninfected controls (Fig. 2A–C). The increased mass of lung was attributed to fluid accumulation, as cell yield following enzymatic lung digestion was not different between infected and uninfected macaques (Fig. 2D). Infectious virus was intermittently recovered at relatively low titer in nasal washes but was consistently isolated from BAL fluid and was present at high titers in all regions of both lungs at necropsy (Table I). Semiquantitative real-time RT-PCR showed uniform and remarkably consistent titers of H5N1 virus in the extremities of upper, middle, and lower regions of both lungs in all animals and revealed significant virus replication in other tissues, most notably lung-draining hilar lymph nodes (Fig. 2E). Virus replication was confirmed by in situ hybridization against H5N1 matrix RNA, which showed intense replication in cells lining most alveoli. Viral RNA was readily detected by in situ hybridization in hilar lymph nodes and duodenum lamina propria, confirming systemic infection (Fig. 2F).
Profound inflammatory response to H5N1 infection in blood and airways of macaques
To begin to define the mechanisms driving lethal disease following aerosol H5N1 inoculation, we first analyzed the cellular response to infection in blood and BAL. Infection produced profound leukopenia at days 1 and 2 PI with loss of CD45+ mononuclear cells and depletion of CD4+ T lymphocytes (Fig. 3A, 3B), consistent with findings in H5N1-infected humans (12, 42). In contrast, there was a significant increase in the number of CD14+CD16+ intermediate monocytes in blood (Fig. 3A, 3B). The cellularity of BAL did not change with infection (Fig. 3C), but the constituent cell populations were dramatically altered. Alveolar macrophages are the predominant cell recovered from BAL in health, and in the macaque, alveolar macrophages are defined as being CD163+CD206+, distinguishing them from CD163+CD206− interstitial macrophages (43). Due to autofluorescence of cells in BAL fluid, matched isotype controls were used with every experiment to ensure that macrophage populations were clearly defined (Supplemental Fig. 1A). Up to 50% of all cells in BAL prior to infection were CD163+CD206+ alveolar macrophages, which were decimated within 2 d of aerosol exposure, coincident with an influx of MHC class II−CD11b+ neutrophils. As expected, CD163+CD206− interstitial macrophages made up an insignificant percentage of cells in BAL before infection, but there was evidence in some animals that infection resulted in a modest increase in interstitial macrophages recovered from BAL (Fig. 3D, 3E). Given that hypercytokinemia is a prominent characteristic of fatal human H5N1 infection (12), we measured a panel of proinflammatory cytokines and chemokines in plasma and BAL using a cytokine bead array before and after infection. There was a transient but profound increase in several proinflammatory factors in plasma, including IL-6, eotaxin, IP-10, and MCP-1, with plasma IL-6 in particular increasing >500-fold within 1 d of infection and then declining (Fig. 3F). There was a massive increase in the concentration of proinflammatory cytokines and chemokines in BAL measured at day 2 PI, with 12 out of 13 analytes significantly increasing. Many proinflammatory cytokines and chemokines in BAL increased several hundred-fold, with IL-6 increasing 1000-fold (Fig. 3G). Notably, the antiviral protein IFN-α increased 2000-fold in BAL following H5N1 infection, reaching concentrations of 10–100 ng/ml at 2 d PI (Fig. 3H).
Reciprocal effects of H5N1 infection on alveolar and interstitial macrophages in lung
We next assessed the cellular response to H5N1 infection in lung cell suspensions, which were enzymatically digested without prior flushing of airways and thus contained both airway and interstitial cells. As with cells in BAL, matched isotype control Abs were used with each stain to ensure that cell populations were appropriately identified (Supplemental Fig. 1B, 1C). The number of AEC, identified by expression of cytokeratin 7/8 (44), recovered from lungs was dramatically decreased as a result of infection in four out five animals, but due to an outlier this effect was not significant (Fig. 4A, 4B). Alveolar macrophages were massively depleted from lung following infection, consistent with findings in BAL, whereas the number of interstitial macrophages, differentiated from the alveolar subset by lack of CD206 expression, increased 4-fold. Lineage−MHC class II+CD206−CD123+ plasmacytoid dendritic cells (DC) constituted a minor population of cells in uninfected lung, which was significantly reduced following infection, whereas the proportion of lineage−MHC class II+ CD206−CD11c+ myeloid DC was unchanged (Fig. 4A, 4B). Interstitial macrophages had significant increases in expression of Ki-67 following infection, revealing recent cell division, whereas the long-lived alveolar macrophages showed no evidence of recent division (Fig. 4C, 4D).
We used immunofluorescence of lung sections to validate and extend flow cytometry findings. Preliminary experiments showed that CD206+ alveolar macrophages all coexpressed CD163, whereas CD163+ macrophages in the lung parenchyma lacked expression of CD206, consistent with flow cytometry data. We therefore for simplicity used Abs to CD206 and CD163 to identify alveolar and interstitial macrophages, respectively, in lung sections. CD206+ alveolar macrophages were readily identified prior to infection but were only rarely present after infection, consistent with findings in BAL and lung suspensions. Conversely, there was a massive influx of CD163+ interstitial macrophages following infection (Fig. 4E). Neutrophils, identified by expression of calprotectin (45), were massively increased in lung sections following infection (Fig. 4E). Immunofluorescence revealed that AEC and to a greater extent neutrophils produced IFN-α following infection, and MCP-1 expression was seen by both cell types. In contrast, infiltrating interstitial macrophages were not found to express IFN-α or MCP-1 (Fig. 4F).
Extensive infection of alveolar macrophages and AEC with loss of epithelial barrier function
We used immunofluorescence to identify the specific cellular targets of H5N1 in the lung using Ab to viral nucleoprotein to identify replicating virus. Nucleoprotein staining substantially colocalized with cytokeratin in alveoli, indicating widespread infection of AEC. Both the highly flattened type I pneumocytes and cuboidal type II pneumocytes extensively expressed virus nucleoprotein (Fig. 5A). Virus replication was detected in alveolar macrophages, although these cells were more difficult to identify because of massive depletion. In contrast, no nucleoprotein staining could be detected in the vast number of infiltrating interstitial macrophages present in infected lung (Fig. 5A). Flow cytometric analysis of lung cell suspensions revealed that ∼20% of all AEC and alveolar macrophages were infected with H5N1, approaching >30% in some animals. Infection of interstitial macrophages averaged <5% across animals (Fig. 5B, 5C). Infection of AEC lead to their apoptosis, as active caspase-3 expression was significantly greater in nucleoprotein-expressing AEC as compared with uninfected AEC in lung cell suspensions from the same animal (Fig. 5D). To determine whether AEC death could have contributed to the leakage of fluid into the alveolar airspace that resulted in acute respiratory distress syndrome following infection, we measured the concentration in BAL of albumin, a large plasma protein that is normally excluded from the airway by an intact epithelial barrier. The concentration of albumin in BAL increased 20-fold in macaques at day 2 PI, and in the three animals surviving to days 4–6 PI, albumin concentration in BAL was 90-fold greater than preinfection levels. This was not a function of hemoconcentration in severely ill animals, as albumin concentrations in plasma did not significantly change with infection (Fig. 5E).
To our knowledge, our study reveals for the first time that inhalation of small-particle aerosols of highly pathogenic avian influenza A (H5N1) virus in cynomolgus macaques induces bilateral pneumonia that rapidly progresses to acute respiratory distress syndrome and death, recapitulating human disease. All previous nonhuman primate studies using newly emerged H5N1 viruses beginning in 2001 have exposed macaques to virus using liquid suspensions instilled into the trachea, often combined with inoculation into the nose, mouth, and eye, an approach that rarely has produced fatal disease (21–26). Studies from three independent groups using the same A/Vietnam/1203/2004 (H5N1) isolate at a similar challenge dose (mean, 7.05 log10 PFU) and in the same species of macaque as used in this study reported fatality in just 2 of 24 (8.3%) animals (21–23, 26).
Avian influenza viruses prefer binding to sialic acid linked to galactose via an α2,3 linkage, and these receptors predominate in the human and nonhuman primate lower respiratory tract rather than the trachea and upper respiratory tract (5, 8). Although H5N1 virus isolates can infect ex vivo cultures of human nasopharyngeal, adenoid, and tonsillar tissues (4), our finding that infectious virus was only intermittently recovered from nasal washes at low titers relative to BAL fluid, despite passage of inhaled virus through the nasopharynx, suggests that the upper respiratory tract is not a major site of replication in vivo. H5N1 infection in macaques was systemic based on high titer of virus in hilar lymph nodes and recovery from peripheral lymph nodes and gastrointestinal tract, although it is possible that the gut was seeded independently through swallowing of virus during aerosol exposure. In human H5N1 cases diarrheal disease is a common presentation and is associated with recovery of virus from feces, consistent with virus replication in the gastrointestinal tract (46).
Close contact with infected domestic poultry is the predominant risk factor for acquiring H5N1 influenza virus infection in humans (47); however, the precise means of avian-to-human transmission remain unknown. Our data indicate that aerosol exposure may be an important source of human H5N1 infections. Sneezing and coughing in H5N1-infected poultry are reported infrequently, as affected animals are typically found comatose or dead with few clinical signs (48). This suggests that respiratory-derived aerosols are an uncommon source of H5N1 virus from infected birds, in contrast to human-to-human transmission of seasonal influenza viruses (49, 50). This may account for the rarity of human infections despite widespread exposure to infected poultry in affected countries (51, 52). H5N1 virus can be recovered from feces and skin of infected birds (48, 53), and manure and skin contribute significantly to fine dust particles (<4 μm diameter) commonly found inside and outside of poultry houses (54), although virus infectivity is rapidly lost in feces upon drying (53). Of greater potential for airborne transmission is inadvertent generation of small-particle aerosols of virus during defeathering and preparing sick or dead poultry for consumption, and this practice is one of the greatest risk factors for human H5N1 infection (47).
Our findings reveal that direct virus-mediated damage to alveoli is a key factor in driving disease severity. Inhalation of aerosolized H5N1 virus resulted in massive infection and apoptosis of AEC, consistent with findings in patients (6) and in vitro (10). Both type I pneumocytes that facilitate gas exchange and type II pneumocytes that secrete surfactant and serve as progenitors for both pneumocyte subsets were affected, accounting for the extreme compromise in respiratory function. Notably, AEC infection was associated with marked protein leakage into the alveolar space, reflecting loss of epithelial barrier function that is a hallmark of the acute respiratory distress syndrome (55). AEC were also found to produce MCP-1, one of the proinflammatory factors that was substantially increased in plasma and BAL fluid and that is correlated with fatal H5N1 infection in humans (12). Indeed, the intense production of proinflammatory cytokines in airways, with 500- to 1000-fold increases in IL-6, eotaxin, IP-10, and IL-8 at 2 d of infection, could reflect the extensive infection of AEC following aerosol infection with H5N1. In comparison, intratracheal infection of macaques with the 1918 influenza virus was associated with modest 25-fold increases in plasma IL-6 by 8 d PI (56).
Inhalation of airborne H5N1 virus also led to up to 30% of alveolar macrophages being infected, resulting in rapid decimation of these cells. Loss of alveolar macrophages would contribute directly to disease through loss of regulatory function (11), and experimental depletion of alveolar macrophages in sublethal influenza virus infection of mice and pigs exacerbates pulmonary inflammation, edema, and fluid leakage (57–59). Loss of alveolar macrophages was offset by massive infiltration of interstitial macrophages, although the function of these newly recruited macrophages remains unclear. Infiltrating interstitial macrophages had recently divided, as distinct from the alveolar macrophages, indicating that they were not simply alveolar macrophages that had downregulated CD206 expression. Macrophage infiltration was associated with significant increases in intermediate CD14+CD16+ monocytes that have an inflammatory function (60) and were likely recruited to lung via increased MCP-1 (61). Infiltrating macrophages contribute to AEC apoptosis in murine influenza models (62, 63), but in our study AEC apoptosis was primarily a consequence of AEC infection. Macrophages and TNF/inducible NO synthase –producing DC have been implicated in cytokine production and immunopathology in murine models (64, 65). A hallmark of these cells in murine influenza is high-level production of TNF-α (65), but recently recruited interstitial macrophages in macaques in our study produced negligible amounts of TNF-α by intracellular cytokine staining (data not shown), and TNF-α was one of the least significant cytokines present in BAL fluid after infection. It is possible that interstitial macrophages are recruited in part to repopulate decimated alveolar macrophages, as has been demonstrated in other macaque models (43). The marginal increase in CD163+CD206− macrophages in BAL at day 2 PI would suggest that this was not a significant effect, although the rapidity of disease could be a factor in limiting transmigration of macrophages into airways.
A notable feature of H5N1 infection in our study was the profound production of IFN-α in the distal airways, which increased nearly 2000-fold by 2 d PI in all animals. IFN-α was produced to a modest extent by AEC but was primarily found in infiltrating neutrophils in animals at necropsy. The principal source of type I IFN in pulmonary infection with RNA viruses is the alveolar macrophage (66), and it is possible that in the early stages of infection prior to their decimation alveolar macrophages accounted for this massive production of IFN-α in airways. Influenza A/Vietnam/1203/2004 (H5N1) infection induces strong induction of types I and III IFN in human cell lines in vitro but results in a paradoxical downregulation of IFN-stimulated genes relative to the less pathogenic 2009 pandemic H1N1 influenza virus, an effect that is linked to H5N1 virus NS1 (67). Similarly, in macaques infected with an identical H5N1 isolate, type I IFN gene expression in lung was robust early in infection but IFN-stimulated gene expression was limited relative to infection with less pathogenic H5N1 isolates, associated with a failure to suppress virus replication (23). These data suggest that H5N1 virus antagonizes IFN-stimulated gene expression despite intense IFN-α induction, facilitating massive virus replication that leads to alveolar destruction.
A limitation of our study is that only one virus strain given at a relatively high challenge dose was used in all animals. It will be important both in terms of public health significance and therapeutic intervention to determine the susceptibility of macaques to a range of doses of aerosolized H5N1 virus and to identify the point at which disease severity is attenuated and survival prolonged. Similarly, comparative studies with a less pathogenic strain that disseminates throughout the lower respiratory tract following aerosolization would allow us to definitively address the mechanism for mortality following H5N1 infection. A good candidate for such studies is 2009 pandemic H1N1, which replicates in lungs and other organs of macaques (68) but does not induce significant disease when delivered by small-particle aerosol (69).
Our data support the notion that rapid initiation of therapies targeting virus replication is indicated in the treatment of H5N1 pneumonia, as opposed to therapies targeting cytokines, consistent with recent murine studies (20). The work establishes aerosolized H5N1 infection of nonhuman primates as the model of choice for evaluating potential antiviral therapies in the treatment of highly pathogenic avian influenza virus infection of humans, as well as testing promising H5N1 vaccine approaches for their ability to prevent fulminant lower respiratory tract disease.
We thank Dana Gillis and Nicole McKinney for veterinary assistance and Michael Kujawa, Aaron Walters, Guoji Wang, Mark Stauffer, and Dana Weber for technical assistance.
This work was supported by funds from the University of Pittsburgh (to S.M.B.-B.) and the University of Georgia (to D.R.P.), as well as by National Institutes of Health Grant AI111598 (to S.J.B. and C.A.W.) and National Institutes of Health Contract HHSN272201400008C (to D.R.P.).
The online version of this article contains supplemental material.
Abbreviations used in this article:
alveolar epithelial cell
Madin–Darby canine kidney
positron emission tomography
The authors have no financial conflicts of interest.