Identification of Oxidative Stress and Toll-like Receptor 4 Signaling as a Key Pathway of Acute Lung Injury

TLR4 Is a Susceptibility Gene for ALI

To identify endogenous genes that control the severity of ALI, we tested inbred mouse strains for their propensity to develop ALI in response to acid aspiration, a model that recapitulates the acute phase of human ARDS (Imai et al., 2005, Nagase et al., 2000). Acid aspiration in mice results in rapid impairment of lung function as determined by increased lung elastance, a measure of the change in pressure achieved per unit change in volume, representing the stiffness of the lungs, decreased blood oxygenation, and alveolar wall thickening, bleeding, inflammatory cell infiltration, and formation of hyaline membranes (Figures S1A–S1C available online). The severity of acid-induced ALI was comparable among various inbred mouse strains with one exception: C3H/HeJ mice exhibited reduced lung elastance. Moreover, C3H/HeJ mice exhibited better blood oxygenation and reduced inflammation (Figures S1A–S1C).

C3H/HeJ mice exhibit resistance to a lipopolysaccharide (LPS) challenge due to a mutation in the intracellular domain of Toll-like receptor 4 (TLR4) (Tlr4^Lps-d) that interferes with TLR4 signaling (Poltorak et al., 1998). To test whether this TLR4 mutation might be involved in ALI severity, we assessed acid-induced ALI in C3H/HeOuJ mice, a C3H-congenic mouse strain that expresses wild-type (WT) TLR4. The resistance to ALI of C3H/HeJ mice was lost in C3H/HeOuJ mice (Figures S1A–S1D). To confirm that TLR4 determines disease severity, we triggered acid-mediated ALI in tlr4^−/− mice (Hoshino et al., 1999). Genetic inactivation of tlr4 markedly attenuated ALI as determined by improved lung elastance ( Figure 1A), reduced edema formation (Figure 1B), and ameliorated histological changes (Figure 1C). Bone marrow chimeras showed that the severity of ALI is controlled by TLR4 expression on hematopoietic cells (Figures 1D–1F). Lung pathology were not affected in tlr3 (Yamamoto et al., 2003) or tlr9 (Hemmi et al., 2000) mutant mice (Figures S2A and S2B). These data identify TLR4 as a susceptibility gene for ALI.

Acute Lung Injury Is Mediated via TLR4-TRIF-TRAF6-NF-κB Signaling

TLR4 stimulation is relayed to cellular responses via different adaptors such as MyD88 and TRIF (Akira et al., 2006). Surprisingly, genetic deletion of myd88 (Adachi et al., 1998) had no apparent effect on the severity of ALI (Figures S2C and S2D). In contrast, trif^−/− mice (Yamamoto et al., 2003) exhibited markedly improved lung function ( Figure 2A), lung pathology (Figure 2B), and lung edema (Figure S3A). TRIF signals either through IKK-ɛ and activation of IRF3 or via TRAF6-mediated NF-κB activation (Sato et al., 2003). To dissect the TLR4-TRIF activation pathway, we conducted experiments using irf3 and traf6 knockout mice. Deletion of irf3 (Sato et al., 2000) did not show a significant change in acid-mediated ALI (Figures S2E and S2F). We next generated myeloid cell-specific traf6 mutant mice using Lys-Cre-mediated deletion of a TRAF6^flox/flox allele (TRAF6^MC-KO). Deletion of TRAF6 was confirmed in cultured bone marrow-derived macrophages (Figure 2C). Inactivation of TRAF6 in TRAF6^MC-KO mice resulted in alleviation of ALI (Figures 2D, 2E, and S3B). To determine whether mutations of TLR4, TRIF, or TRAF6 affect NF-κB activation, we performed in situ signaling experiments to detect nuclear accumulation of Ser276-phosphorylated NF-κBp65 indicative of NF-κB activation (Vermeulen et al., 2002). Whereas acid aspiration triggered NF-κBp65 activation in WT and myd88^−/− mice, much less nuclear NF-κBp65 translocation occurred in lungs from tlr4^−/−, trif^−/−, or TRAF6^MC-KO mice (Figure 2F). NF-κB activation was largely restricted to macrophages. Taken together, our genetic results identify TLR4-TRIF-TRAF6-NF-κB as a key signaling pathway that links acid injury to the severity of ALI.

IL-6 Mediates Acute Lung Injury

Local cytokine production at the site of injury is one of the hallmarks of ALI (Martin, 1999). The molecular basis of how such cytokine storms are mediated by local injury and whether such injury-released cytokines actually contribute to disease severity of ARDS are still unknown. To examine whether TLR4 signaling may be involved in local cytokine, we performed multiple cytokine arrays from normal lungs and lung tissue following acid aspiration. The cytokines IL-6, IL-1β, and KC (IL-8) were upregulated by acid treatment in WT mice, but we failed to observe upregulation of TNFα, IL-2, IFN-γ, IL-1α, IL-12p40, IL-12p70, IL-17, RANTES, IL-9, Eotaxin, or MIP-1β (Figure S4 and data not shown). IL-3, IL-4, IL-5, IL-13, G-CSF, GM-CSF, MCP-1, and MIP-1α were below the detection limit of our assay. IL-10 negatively correlated with the severity of ALI (Figure S4E). In particular, IL-6 levels closely reflected the severity of ALI in resistant tlr4^−/−, trif^−/−, and TRAF6^MC-KO mice and injury-prone mice myd88^−/− and irf3^−/− mice (Figure 2G). To test if IL-6 is indeed involved in the pathogenesis of ALI, we studied il-6^−/− mice (Kopf et al., 1994). In response to acid aspiration, il-6^−/− mice exhibited significant improvement of lung function (Figure 2H), lung edema formation (Figure S3C), and decreased lung pathologies (Figure 2I). Although these data do not exclude the importance of other cytokines, our genetic results show that loss of IL-6 alleviates the severity of ALI (Figure 2J).

Oxidative Stress and Formation of Oxidized Phospholipids in ALI

Since we used acid aspiration to trigger ALI, a key question emerged: what triggers TLR4 under such injury conditions? To rule out the possibility of LPS contamination, we first treated mice with polymyxin B to neutralize LPS (Jacobs and Morrison, 1977). Polymyxin B did not alter disease severity in acid-induced ALI (Figure S5A). Moreover, in line with previous data from our group (Imai et al., 2005) and others (Nagase et al., 2000), instillation of LPS alone into lungs does not worsen lung function in our acute lung failure model (Figure S5B). We next stimulated alveolar macrophages from WT, tlr4^−/−, myd88^−/−, or trif^−/− mice with bronchoalveolar lavage (BAL) fluid from normal mice (BAL control), BAL fluid from lungs of acid-treated WT mice (BAL acid), or LPS. In line with our genetic data, BAL fluid-induced IL-6 production was mediated via TLR4 and TRIF, but not MyD88 ( Figure 3A). By contrast, LPS-induced IL-6 production in alveolar macrophages was dependent on TLR4, MyD88, and TRIF (Figure 3A). Thus, other TLR4-activating agents rather than LPS are generated during the course of ALI.

Since the lung air-liquid interface is exposed to the aerobic environment and therefore susceptible to oxidation, we speculated that ALI could result from the local generation of reactive oxygen species (ROS) and the subsequent formation of oxidized phospholipids (OxPLs). We indeed observed induction of ROS in alveolar macrophages following in vivo acid challenge (Figure 3B). Acid treatment also triggered enhanced TLR4 surface expression (Figure 3C). To demonstrate the role of OxPLs in vivo we used the well-characterized mAb EO6 (Friedman et al., 2002). This mAb specifically detects the phosphocholine (PC) headgroup of phospholipids in which the sn2 side chain has been oxidized but does not recognize native nonoxidized PLs. We observed large amounts of EO6-detectable OxPLs in the BAL fluid of acid-challenged animals (Figure S5C). In situ immunohistochemistry confirmed the generation of EO6-detectable OxPLs in alveolar macrophages as well as inflammatory exudates lining the injured air spaces (Figure 3D). In addition, we observed formation of the lipid peroxidation breakdown product malondialdehyde detected by the specific mAb MDA2 (Palinski et al., 1996) (Figure S5D), further confirming that lipid oxidation does occur in diseased lungs. These data show that acid aspiration triggers the oxidative stress machinery and the generation of OxPLs in lung.

OxPLs Induce Cytokine Production and Acute Lung Injury via TLR4

Minimally modified low-density lipoprotein (LDL) has been previously shown to mediate macrophage activation (Miller et al., 2003a, Miller et al., 2005) and to modulate the severity of atherosclerosis and inflammatory responses via TLR4 (Miller et al., 2003a, Michelsen et al., 2004). To address whether OxPLs generated during ALI can activate innate immune responses, we stimulated WT and tlr4^−/− alveolar macrophages with BAL fluid from normal (BAL control) and diseased lungs (BAL acid). BAL fluid from mice with ALI, which contains OxPLs (Figure S5C), induced large amounts of IL-6 production in WT alveolar macrophages. Neutralization of OxPLs by the specific mAb EO6 attenuated BAL fluid-induced IL-6 production (Figure 3E). BAL fluid did not trigger IL-6 release in tlr4^−/− alveolar macrophages (Figure 3E). Similarly, BAL fluid from diseased mice induced IL-6 production in peritoneal (Figure S6A) and bone marrow macrophages (Figure S6B), which was reverted by EO6 Ab treatment. Induction of IL-6 was again dependent on TLR4 expression (Figures S6A and S6B). OxPLs were also observed in acid-treated trl4^−/− mice (Figure S6C).

The major source of PLs in the lung is surfactant, which forms a film at the alveolar air-water interface and reduces surface tension. Surfactant contains 80%–90% PLs including unsaturated phosphatidylcholine that can be oxidized (Rodriguez Capote et al., 2003, Veldhuizen et al., 1998). In vitro-oxidized surfactant PLs contain EO6-detectable OxPLs (Figure S6D). Addition of purified synthetically oxidized surfactant PLs induced IL-6 production in a dose-dependent manner in WT peritoneal macrophages (Figure 3F). IL-6 production in response to surfactant OxPLs was decreased by the neutralizing mAb EO6 (data not shown) and dependent on TLR4 expression (Figure 3F). In line with previous results (Bailey et al., 2004), in vivo administration of nonoxidized PLs induces slightly impaired lung function (Figure 3G). Intratracheal administration of OxPLs markedly worsened lung function in normal mice (Figure 3G). In a second experimental model, we removed endogenous surfactant via saline lavages, which strongly impairs lung function; we followed this with intratracheal replacement with nonoxidized or oxidized surfactant PLs. In this scenario, challenge with OxPLs resulted in much more severe impairment of lung function and all experimental animals died by 2 hr, whereas mice that received control, nonmodified surfactant PLs stabilized their lung function (Figure S7A). Lung function was comparable among WT and tlr4^−/− mice following administration of nonoxidized surfactant PLs. Loss of TLR4 expression markedly alleviated the severe lung failure phenotype following OxPLs treatment (Figure 3H). Moreover, administration of OxPLs induced increased levels of IL-6 in the lungs of WT control but not in those of tlr4^−/− mice (Figure S7B). These data indicate that synthetically oxidized surfactant PLs can trigger cytokine production in macrophages and, importantly, acute lung injury in vivo via TLR4.

OxPAPC Triggers Acute Lung Injury

We next attempted to identify specific OxPLs that can trigger ALI via TLR4. The mAb EO6 we used for the detection of OxPLs has been previously reported to specifically bind PC-containing OxPLs such as oxidized 1-palmitoyl-2-arachidonoyl-phosphatidylcholine (OxPAPC) (Figure S7C). We confirmed that the mAB EO6 indeed detects OxPAPC (Figure S8A). To directly test whether OxPAPC can trigger cytokine production via TLR4, we stimulated lung tissue macrophages with OxPAPC and nonoxidized PAPC. Compared to PAPC, OxPAPC induced higher IL-6 levels in WT macrophages, and these levels were attenuated by the neutralizing mAb EO6 ( Figure 4A). OxPAPC-induced IL-6 production was decreased in tlr4^−/− and trif^−/− lung macrophages, and this residual IL-6 production was not affected by the mAb EO6 (Figure 4B). Similar to our in vivo experiments, OxPAPC could still trigger IL-6 production in myd88^−/− lung tissue macrophages. LPS-induced IL-6 production in lung macrophages was dependent on TLR4, TRIF, and MyD88 (Figure 4B). In addition, OxPAPC induced higher IL-6 production than nonoxidized PAPC in alveolar macrophages isolated by lung lavage (Figure S8B). OxPAPC-induced IL-6 production was decreased in tlr4^−/− alveolar macrophages (Figure S8C). Similar to lung macrophages, OxPAPC triggered IL-6 production in peritoneal macrophages dependent on TLR4 and TRIF but not MyD88 (Figure S8D). Thus, OxPAPC can trigger IL-6 production in lung and peritoneal macrophages via TLR4-TRIF signaling. Importantly, intratracheal administration of OxPAPC induced marked deterioration of lung function in WT mice (Figure 4C). In vivo OxPAPC, but not PAPC, challenges triggered IκBα degradation indicative of an activated NF-κB pathway (Figure 4D). Moreover, lung function (Figure 4E) and IL-6 production (Figure S8E) were markedly alleviated in tlr4^−/− and trif^−/− but not myd88^−/− mice. These data indicate that OxPAPC can trigger ALI via TLR4 and TRIF.

Inactivated H5N1 Avian Influenza Virus Can Induce OxPL and ALI in Mice

Patients who died of H5N1 avian influenza or Spanish influenza developed ARDS, and extremely high levels of cytokines have been observed in patients and animals infected with these viruses (Beigel et al., 2005, Tumpey et al., 2005). Our data so far showed that acid aspiration triggers production of OxPLs that can augment the severity of ALI and cytokine production via TLR4. We wanted to extend these findings to H5N1-mediated ALI. Inactivated influenza A virus (McKinney et al., 2003) or inactivated respiratory syncytial virus (Haeberle et al., 2002) can induce rapid activation of NF-κB and/or proinflammatory cytokine production in macrophages. The virulence of the 1918 Spanish influenza virus was largely determined by its hemaglutinin (HA), and the recombinant viruses expressing the 1918 viral HA induce high IL-6 production and severe ALI in vivo (Kobasa et al., 2004). Based on these reports, we tested whether an inactivated H5N1 virus where HA activity is conserved might induce OxPL production and ALI in vivo via TLR4. Since H5N1 subtype influenza A viruses are more potent inducers of cytokines in macrophages than H1N1 subtype influenza (Cheung et al., 2002), we used H1N1 as a control.

Pulmonary challenge of inactivated H5N1 avian influenza virus induced rapid impairment of lung function as determined by increased lung elastance ( Figure 5A), severe edema formation as assessed by wet/dry lung weight ratios (Figure 5B), and pathological changes such as alveolar wall thickening, bleeding, or the accumulation of inflammatory cells (Figure S9A). Administration of inactivated H1N1 virus had only very mild effects on lung function and lung histology (Figures 5A, 5B, and S9A). In line with a previous study (van Riel et al., 2006), we observed influenza A virus nuclear protein-positive cells in the lower respiratory tract (LRT) in mice treated with inactivated H5N1 virus (Figure 5C). By contrast, we only found very few virus nuclear protein-positive cells in the LRT of H1N1-treated mice (Figure 5C). Similar to acid treatment, we found higher production of ROS in alveolar macrophages isolated from mice treated with inactivated H5N1 avian influenza virus (Figure 5D). In addition, TLR4 surface expression was increased in alveolar macrophages from mice treated with inactivated H5N1 (Figure 5E). In vivo challenge of mice with H1N1 did not result in detectable ROS production or upregulation of TLR4 cell-surface expression (Figures 5D and 5E). Moreover, we detected OxPLs in the BAL fluid of animals challenged with inactivated H5N1 (Figure S9B). Immunohistochemistry confirmed enhanced generation of EO6-detectable OxPLs at the sites of injury (Figure S9C). We extended the analysis of OxPLs to live H5N1 infections. Similar to inactivated H5N1 (Figure 5C), influenza virus A protein was detected in the LRT of mice infected with live H5N1 avian influenza virus (Figure 5F). Importantly, infection with live H5N1 avian influenza virus resulted in the formation of EO6-detectable OxPLs in mouse lung (Figure 5F). Thus, challenge of mice with H5N1 avian influenza can trigger OxPLs and rapid onset ALI in vivo.

Inactivated H5N1-Induced ALI Is Attenuated in tlr4 and trif Mutant Mice

To test whether pulmonary exposure of inactivated H5N1 viruses affects the severity of ALI via the TLR4-TRIF pathway, we analyzed ALI in tlr4^−/− and trif^−/− mice. The tlr4^−/− mice, treated with the inactivated H5N1 virus, showed significantly improved lung elastance ( Figure 6A), pulmonary edema formation (Figure 6B), and alleviated lung pathologies (Figure 6C) compared with control WT mice. Moreover, trif^−/− mice, treated with the inactivated H5N1 virus, showed improved lung elastance (Figure 6D), pulmonary edema formation (Figure S9D), and alleviated lung pathologies (Figure 6E). NF-κBp65 activation in macrophages was also impaired in lungs from tlr4^−/− and trif^−/− mice (Figures 6C and 6E,). In addition, BAL fluid obtained from H5N1-challenged mice induced high levels of IL-6 in WT but not tlr4^−/− alveolar macrophages. Neutralization of OxPLs using the mAb EO6 attenuated IL-6 production in WT but not tlr4^−/− alveolar macrophages (Figure 6F). Thus, cytokine production in alveolar macrophages and ALI in vivo are attenuated in tlr4 and trif mutant mice in response to challenge with inactivated H5N1 avian influenza virus.

H5N1 Avian Influenza Virus Triggers Oxidative Stress in Humans

Next, we examined whether inactivated H5N1 avian influenza viruses can also trigger the oxidative stress machinery in humans. Stimulation of peripheral blood mononuclear cells (PBMCs) with inactivated H5N1 viruses indeed resulted in generation of ROS ( Figure 7A). By contrast, challenge with inactivated H1N1 resulted in much less ROS production in human PBMCs. Moreover, FACS staining showed that TLR4 surface expression was upregulated in PBMCs treated with inactivated H5N1 compared to H1N1 (Figures 7B and S10A). Whereas TLR4 protein is distributed throughout the cytoplasm in human PBMCs under control conditions, challenge of PBMCs with inactivated H5N1 virus induced relocalization and cell membrane association of TLR4 (Figure 7C). Thus, inactivated H5N1 viruses can induce ROS formation and TLR4 surface expression on primary human monocytes.

Loss of NCF1 Improves the Severity of ALI

Our data so far showed that inactivated H5N1 viruses can trigger ROS formation in mouse alveolar macrophages and human monocytes. We therefore want to test whether such ROS generation is indeed involved in disease pathogenesis of H5N1 influenza virus-mediated ALI. To address this, we analyzed lung injury in WT and ncf1 (neutrophil cytosolic factor 1) mutant mice (Hultqvist et al., 2004). Ncf1 is a key component of the NADPH oxidase complex required for oxidative burst and ROS formation (Hultqvist et al., 2004). This ncf1 mutation affects splicing leading to a near absence of Ncf1 protein (Huang et al., 2000). Interestingly, the ncf1 mutant mice treated with inactivated H5N1 virus showed significantly improved lung elastance (Figure 7D), pulmonary edema formation (Figure S10B), and alleviated lung pathologies (Figure 7E) compared with control mice on the same genetic background. Importantly, in H5N1-challenged ncf1^−/− mice we failed to detect oxidized PL formation (Figure 7E). Thus, genetic impairment of ROS production results in reduced formation of OxPLs and significantly improved lung functions in response to H5N1 challenge.

Multiple Lethal Lung Pathogens Trigger Formation of OxPLs in Different Species

Finally, to test whether OxPLs are generated during actual H5N1 influenza virus infections in humans, we analyzed lung samples from two patients who had developed ARDS following H5N1 avian influenza virus infections. Consistent with our mouse acid aspiration and H5N1 models, we detected massive formation of OxPLs localized at the inflammatory exudates lining the injured air spaces as well as alveolar macrophages in the lungs of both H5N1-infected patients (Figure 7F). Lungs from patients that died of extrapulmonary diseases did not show any detectable OxPL formation (Figure S10C). Thus, H5N1 infections in humans result in the local activation of the oxidative stress machinery and OxPL formation in the lung.

To examine whether OxPL production is a general feature of lethal lung injury that can be seen in different pathogen infections and across wider species barriers, we analyzed lung samples from nine individuals who had developed ARDS following SARS infections. In all SARS-infected human cases tested, we observed marked production of OxPLs in the inflammatory exudates lining the injured air spaces, pneumocytes, as well as alveolar macrophages (Figure 7F and data not shown). To extend these findings to additional species and pathogens, we analyzed lung samples from animals infected with potential bioterrorism agents. We observed formation of OxPLs in Anthrax-infected Rhesus monkeys, Anthrax-infected rabbits, Monkey Pox-infected Cynomolgus monkeys (Macaca fascicularis), as well as Yersinia pestis-infected Cynomolgus monkeys that developed lung plague (Figure 7G). By contrast, we could not detect elevated OxPLs in Smallpox or Marburg virus-infected Cynomolgus monkeys, nor in monkeys infected with Western Equine Encephalitis (WEE) or Eastern Equine Encephalitis (EEE) viruses, diseases that did not cause severe lung injury (data not shown). Thus, in multiple species, infections with various lethal lung pathogens such as H5N1 avian influenza virus, SARS-coronavirus, Anthrax, Y. pestis, or Monkey Pox virus can trigger formation of OxPLs in the lung. These results should be extended to additional pathogens and species.

Clinical Patient Profiles

Lung samples from two H5N1- and nine SARS-infected humans were used.

Animal Infections

C57BL6 mice were infected intranasally with live H5N1 avian influenza virus (strain A/HK/483/97) and euthanized for tissue sampling on day 4. Lethal lung infections with Anthrax, Y. pestis, and Monkey Pox virus, as well as infections with Smallpox virus, Marburg virus, WEE, and EEE and lung sampling were performed at the US Army Medical Research Institute of Infectious Diseases.

Mutant Mice

tlr4, tlr9, myd88, tlr3, trif, irf3, il-6, and ncf1 mutant mice have been previously described. Mice harboring a floxed TRAF6 allele will be reported in the future and were crossed onto LysCre mice (Clausen et al., 1999). Only sex-, age-, and background-matched mice were used. Basal lung function and lung structure were comparable among all the mice tested.

Acid Aspiration-Induced ALI in Mice

After intratracheal instillation of HCl (pH = 1.5; 2 ml/kg), animals were ventilated for 3 hr. Total positive end expiratory pressure (PEEP_t) and plateau pressure (P _plat) were measured at the end of expiratory and inspiratory occlusion. Elastance was calculated as (P _plat − PEEP_t) divided by tidal volume (V _T) every 30 min during the ventilation periods.

Bone Marrow Transplantation

Recipient mice were irradiated and 1 × 10⁶ donor bone marrow cells were delivered i.v. through the tail vein. Chimeras were used for experiments 12 weeks after transplantation.

Preparation of Oxidized Surfactant PLs and In Vivo Treatments

In vitro oxidation of surfactant PLs was performed according to previously established conditions (Rodriguez Capote et al., 2003). The in vivo impact of nonoxidized surfactant PLs and oxidized surfactant PLs was tested in normal, healthy lung as well as saline-lavaged lungs to remove surfactant.

Polymyxin B, LPS, and OxPAPC Treatment in Mice In Vivo

To neutralize LPS, mice received polymyxin B or control vehicle intratracheally. LPS (E. coli O111:B4; 0.5 μg/g body weight), OxPAPC (20 μg/g body weight), or control PAPC (20 μg/g body weight) were instilled intratracheally. Animals were then ventilated for 3 hr and pulmonary elastance determined.

H5N1 Avian Influenza Isolates and H5N1-Induced Lung Injury in Mice

Inactivated avian influenza subtypes (H1N1 and A/ck/Yamaguchi/7/04 H5N1) were obtained from the Istituto Zooprofilattico Sperimentale delle Venezie. Vehicle and inactivated H1N1 or H5N1 viruses were instilled intratracheally. Animals were then ventilated for 5 hr and elastance was calculated every 30 min.

Assessment of Blood Oxygenation and Pulmonary Edema

Partial pressure of arterial oxygen (PaO₂) was measured to assess arterial blood oxygenation as an indicator for respiratory failure. To assess pulmonary edemas, the lung wet/dry weight ratios were calculated.

Histology, In Situ NF-κB Detection, and Immunohistochemistry

Lung tissue was fixed in 4% buffered formalin. For histological analysis, sections were stained with hematoxylin and eosin (H&E). For detection of activated NFκB, paraffin sections were incubated with a polyclonal anti-phospho-NFκBp65 (Ser276) antibody. Immunohistochemistry to detect oxidized phospholipids (OxPLs) was performed using the mAb EO6 (kindly provided by J.L. Witztum). Immunohistochemistry for influenza A virus nucleoprotein was performed using the HB 65 (EVL anti-influenza NP, subtype A) antibody.

Multiple Cytokine Expression Analyses and Cytokine ELISA

Frozen lung tissues were homogenized in cell lysis buffer, and supernatants were assayed using multiplex cytokine arrays. For ELISA analyses, supernatants of homogenized lung tissue were assayed using specific kits.

Isolation and Activation of Macrophages

Bone marrow macrophages were isolated from femurs of 8- to 10-week-old mice. Peritoneal macrophages were obtained by injecting mice with thioglycollate, followed by peritoneal lavage. Alveolar macrophages were obtained from mice by bronchoalveolar lavage (BAL). Lung tissue macrophages were prepared using collagenase and DNase. For measurement of IL-6, macrophages were plated on 96-well plates and stimulated in vitro for 24 hr.

Detection of ROS and TLR4

Mouse alveolar macrophages were obtained from mice treated with vehicle, acid, or inactivated H5N1. Human PBMC were isolated with Ficoll-Hypaque and treated with inactivated H5N1 or H1N1 virus. For ROS detection, cells were incubated with DCF. For mouse TLR4 surface staining, cells were double stained for CD11c and TLR4. For human TLR4 detection, cells were double stained for TLR4 and CD14. ROS and TLR4 expression were assessed by FACS. For TLR4 immunolocalization, PBMCs were fixed, permeabilized, and incubated with anti-TLR4.

Statistical Analyses

All data are shown as mean ± standard error of the mean (SEM). Measurements at single time points were analyzed by ANOVA and, if significant, further analyzed by a two-tailed t test. Time courses were analyzed by repeated-measurements (mixed model) ANOVA with Bonferroni post-tests.