Fatal Outcome of Pandemic H1N1 2009 Influenza Virus Infection Is Associated with Immunopathology and Impaired Lung Repair, Not Enhanced Viral Burden, in Pregnant Mice

Abstract

Pandemic A (H1N1) 2009 influenza virus (pH1N1) infection in pregnant women can be severe. The mechanisms that affect infection outcome in this population are not well understood. To address this, pregnant and nonpregnant BALB/c mice were inoculated with the wild-type pH1N1 strain A/California/04/09. To determine whether innate immune responses are associated with severe infection, we measured the innate cells trafficking into the lungs of pregnant versus nonpregnant animals. Increased infiltration of pulmonary neutrophils and macrophages strongly correlated with an elevated mortality in pregnant mice. In agreement with this, the product of nitric oxide (nitrite) and several cytokines associated with recruitment and/or function of these cells were increased in the lungs of pregnant animals. Surprisingly, increased mortality in pregnant mice was not associated with higher virus load because equivalent virus titers and immunohistochemical staining were observed in the nasal cavities or lungs of all mice. To determine whether exacerbated inflammatory responses and elevated cellularity resulted in lung injury, epithelial regeneration was measured. The lungs of pregnant mice exhibited reduced epithelial regeneration, suggesting impaired lung repair. Despite these immunologic alterations, pregnant animals demonstrated equivalent percentages of pulmonary influenza virus-specific CD8⁺ T lymphocytes, although they displayed elevated levels of T-regulator lymphocytes (Tregs) in the lung. Also, pregnant mice mounted equal antibody titers in response to virus or immunization with a monovalent inactivated pH1N1 A/California/07/09 vaccine. Therefore, immunopathology likely caused by elevated cellular recruitment is an implicated mechanism of severe pH1N1 infection in pregnant mice.

INTRODUCTION

In most cases, the pandemic H1N1 2009 influenza virus (pH1N1) strain is self-limiting and mirrors infection by commonly circulating seasonal influenza virus strains. However, pregnant and postpartum women are at greater risk for severe clinical manifestations (20, 44, 47, 73). In one study, the Centers for Disease Control and Prevention reported a death rate of 20% in pregnant women in the United States who were hospitalized with pH1N1-related symptoms (72). Among these deaths, 64% occurred within the third trimester of pregnancy. Serious influenza-induced complications in pregnant women are not a new phenomenon. In sharp contrast to the general population, pregnant women had disproportionately high rates of hospital admission and mortality during the previous influenza pandemics: the 1918-1919 Spanish, 1957-1958 Asian, and 1968-1969 Hong Kong pandemics (50, 65). The mechanism for the greater severity of pH1N1 and pandemic influenza viruses during pregnancy remains elusive.

Pregnancy is associated with immunological suppression (2) and increased susceptibility to pathogens (4, 45, 64). Typically, the lack of appropriate or sufficient adaptive immune responses is associated with the severity of influenza infection. The only known clinical study linking adaptive immunity to severe pH1N1 infection during pregnancy is a deficiency in serum IgG2 (26). In that report, severely infected pregnant women displayed decreased levels of serum IgG2, suggesting a role for the IgG subclasses in the outcome of pH1N1 infection. Alterations in the immunologic responses that follow early infection also determine disease outcome.

The secretion of cytokines and chemokines is critical in limiting influenza replication (14, 24, 25, 36). Interestingly, several studies support the idea that dysregulation of innate immune responses and elevation of the monocyte recruitment factor, monocyte chemoattractant protein 1 (MCP-1), are a hallmark feature of severe influenza infection in humans (29) and mice (1, 77). Alterations in innate immune responses to infection during pregnancy include decreased cellular responses (42), changes in cytokine production (79), and skewing of the T-helper 1 and 2 (Th1/Th2) balance (80). Some evidence points to changes in cytokine responses associating with severe pH1N1 infection in pregnant mice (12). Excessive recruitment of pulmonary macrophages and neutrophils is positively associated with severe infection with pandemic 1918 H1N1 influenza virus (62). The most commonly reported symptoms associated with enhanced morbidity in pregnant women are severe pulmonary disease (32, 57), demonstrating that lung involvement is an important feature that determines disease severity. However, it is unclear what cell types are being recruited to this site of infection and whether all or any of these factors are responsible for the severity of pH1N1 infection during pregnancy. The impact of these infiltrates on the severity of pH1N1 infection in pregnant animals is unknown. However, several models of severe infection show a role for infiltrates at the site of infection in causing severe immunopathology (34, 73, 78, 82, 83). Furthermore, evidence suggest that the lethality of influenza infections occurs as a result of immune-based tissue injury rather than exacerbated viral burden (31, 71, 74). This suggests that the host response may contribute to disease severity through immune pathology-based mechanisms.

We used BALB/c mice during the middle to late gestational period to mirror the trimester of pregnancy most associated with severe pH1N1 infection in humans (72). Unlike what was previously reported (12), in our model, lethal infection in pregnant mice was independent of increased viral burden but was associated with elevated pulmonary chemoattractants such as MCP-1 and enhanced numbers of macrophages and neutrophils. The recruitment of these cell populations was further characterized by an increase in the nitric oxide (NO) metabolite nitrite in the lungs of pregnant animals. Also, these mediators were associated with decreased regeneration of the epithelial lining of the pulmonary airways. Our findings imply that immunopathology-based lung injury is an associated mechanism of severe pH1N1 infection in pregnant mice.

Little is known about the immunogenicity and efficacy of pandemic H1N1 2009 vaccines in pregnant women (8, 81). Despite the immunological alterations and reduced epithelial regeneration associated with increased mortality, we found that pregnant mice mounted antibody responses to pH1N1 infection or monovalent inactivated pandemic vaccine similar to nonpregnant controls. This finding reinforces the public health message that pregnant women should be a high-priority group targeted for vaccination.

MATERIALS AND METHODS

Mice.

Eight- to 9-week-old pregnant (day 13 gestation) and age-matched nonpregnant BALB/c female mice were obtained from Charles River Laboratories (Wilmington, MA) and housed in the Animal Resource Center specific-pathogen-free (SPF) facility at St. Jude Children's Research Hospital (SJCRH). All experimental procedures conducted with mice were approved by SJCRH Institutional Animal Care and Use Committee.

Virus propagation and inoculation into mice.

Wild-type (wt) human influenza viruses, including the seasonal virus A/Brisbane/59/07 H1N1 (Brisbane) and pandemic virus A/California/04/09 H1N1 (pH1N1), were obtained from the World Health Organization influenza collaborating laboratories and propagated in the allantoic cavities of 10-day-old embryonic chicken eggs. Brisbane was used only in vaccine studies. The infection titers of Brisbane and pH1N1 were determined by using the Reed and Muench method (66) and expressed as the log₁₀ 50% egg infectious dose (EID₅₀) of fluid. Mice were intranasally inoculated with 30 μl of either 10⁴ EID₅₀ of Brisbane or 10⁵ to 10⁶ EID₅₀ of pH1N1 diluted in phosphate-buffered saline (PBS) or with PBS alone after sedation with isoflurane. The 50% lethal dose 50 (LD₅₀) of pH1N1 in pregnant and nonpregnant animals was determined as previously described (66). All inoculated animals were monitored daily.

Virus titrations.

Lung, liver, spleen, kidneys, and brains of 5 mice per group were harvested 3, 5, and either 7 or 9 days postinfection (dpi) and homogenized as described previously (9). The titer of virus in each sample was calculated by the Reed and Muench method (66) and expressed as the 50% tissue culture infective dose (TCID₅₀).

Histopathology.

On 3, 5, and either 7 or 9 dpi, the lungs, livers, spleens, kidneys, brains, and nasal cavities were harvested from 4 or 5 mice per group immediately after they were euthanized with Avertin (2,2,2-tribromoethanol; Sigma-Aldrich, St. Louis, MO). Lungs (sufflated prior to fixation), decalcified nasal cavities, livers, spleens, kidneys, and brains were fixed with 10% neutral buffered formalin overnight. Tissues were processed for hematoxylin and eosin (H&E) or immunohistochemical (IHC) staining as described previously (49). IHC slides were incubated with one of the following primary antibodies: rat anti-mouse neutrophils (Ly6g) (Cell Sciences, Canton, MA), rat antimacrophage (F4/80) (Caltag, Burlingame, CA), goat anti-influenza H1N1 nucleoprotein (NP) (US Biologicals, Swampscott, MA), or rabbit anti-arginase-1 (Santa Cruz, Santa Cruz, CA). All staining was done by the Veterinary Pathology Core Laboratory at SJCRH; IHC staining and epithelial regeneration were analyzed by a veterinary pathologist (SJCRH). Each histopathological slide shown in this article is representative of tissue sections from 4 or 5 mice per group.

Cytokine/chemokine quantification.

Bronchoalveolar lavage (BAL) supernatants were harvested from 4 or 5 mice per group on 3, 4, 5, 6, 7, 8, and 9 dpi, and the pg/ml concentration of 32 mouse cytokines/chemokines was determined using the multiplex bead system according to the manufacturer's instructions (Millipore Corporation, Billerica, MA). In separate studies, sera were collected from blood 3, 5, 7, and 9 dpi for cytokine/chemokine analysis. Cytokines/chemokines were measured using the Millipore Luminex 200 instrument (Luminex Corporation, Austin, TX) and quantified with Luminex xPonent version 3.1 instrumentation software (Luminex Corporation).

Nitrite measurement.

Nitrite, which is the product of nitric oxide, was measured by the Griess method as described previously (18, 68). Briefly, nitrite standard solutions were prepared from a 10 mM stock, and 100 μl was added to duplicate wells using a 96-well plate. On the same plate, 100-μl volumes of the BAL supernatant and appropriate sample blank were added, and then 100 μl of the Griess reagent mixture (reagents A and B in equal parts) was added to all wells that convert nitrite into a purple azo compound. Absorbance was read within 5 min at 540 nm using a plate reader. The concentration of nitrite was expressed as a μM value for each sample.

Flow cytometric analysis.

Single-cell suspensions of BAL fluid were collected from 4 or 5 mice per group at 3, 4, 5, 6, 7, 8, and 9 dpi. Mediastinal lymph nodes (MLN) and spleens were harvested 9 dpi. After nonspecific Fc receptor binding was blocked using purified anti-CD16/CD32 (eBioscience, San Diego, CA), cells were incubated in the dark and stained with antibodies. For macrophage, neutrophil, and dendritic cell (DC) analysis, cells were stained as previously described (1). Live cells were first gated using CD45-allophycocyanin (APC) (BD Biosciences), to exclude erythrocytes and other unwanted cells, and then gated using the following parameters: for macrophages, CD11b _high, Ly6g⁻, and high major histocompatibility complex class II [MHC-II_high]; for neutrophils, CD11b_high, Ly6g _high, and MHC-II⁻; and for DCs, CD11c _high, Ly6g_low, and MHC-II_high. For T-regulator lymphocyte (Treg) analysis, cells were stained with anti-CD4-fluorescein isothiocyanate (FITC) (BD Biosciences, Rockville, MD) and anti-CD25-APC (BD Biosciences) for 30 min at 4°C and then fixed/permeabilized (Cytofix/Cytoperm kit; BD Biosciences) before staining with anti-FoxP3-phycoerythrin (PE) (BD Biosciences). Tregs were identified by gating all live cells that were CD4⁺ CD25⁺ and from this population gating on all FoxP3⁺ CD25⁺ cells. For influenza virus-specific CD8⁺ T lymphocyte analysis, cells were stained for 1 h at room temperature with PE-labeled tetramers specific for the influenza H2^d epitope, NP_144-155 (75), prior to staining with anti-CD8-APC (BD Biosciences) and anti-CD69 PE-Cy7 (BD Biosciences) for 30 min at 4°C. Tetramer-positive CD8⁺ T lymphocytes were identified by gating all live cells that were CD8⁺ and from this population gating on all tetramer⁺ CD8⁺ T lymphocytes. Activated CD8⁺ T lymphocytes were determined by gating on CD69⁺ CD8⁺ T lymphocytes from the total tetramer⁺ CD8⁺ T cell population. Stained samples (minus Treg cells) were fixed with 2 to 4% paraformaldehyde and acquired on either a FACScan or BD FACS Canto II (BD Biosciences). All samples were analyzed using FlowJo software (Treestar, Ashland, OR). Viable cells were counted using C-Chip disposable hemocytometers (Neubauer improved; Incyto, Chungnam-do, Republic of Korea).

Vaccination.

To mirror the immune response to vaccine in humans previously exposed to seasonal influenza A viruses, female mice (n = 6/group) were inoculated intranasally with 10⁴ EID₅₀ of Brisbane. Eighteen days later, mice were allowed to mate with male mice (or not mated) for 3 days. Female mice were then immunized intramuscularly with 6 μg monovalent inactivated H1N1 A/California/07/09 vaccine (or PBS alone) on days 21 and 27. Serum samples were collected 3 weeks after the second vaccination to determine antibody titers.

Antibody detection.

To measure antibody titers in sera in response to pH1N1, Brisbane, or vaccine, blood was collected from mice via the retro-orbital method. Inhibition of virus was determined using the hemagglutination inhibition (HI) assay (9). To test for pH1N1-specific antibodies, sera were collected 2 weeks after infection with a sublethal dose or 9 dpi after a lethal infection with wt A/California/04/09. Virus-specific Ig or IgG subclass (IgG1, IgG2a, IgG2b, IgG2c, and IgG3) titers were determined by enzyme-linked immunosorbent assay (ELISA) (48). Antibodies were detected using alkaline phosphatase-conjugated goat anti-mouse antibodies (Southern Biotech, Birmingham, AL).

Statistical and image analysis.

Significant differences in Kaplan-Meier survival curves were determined by using the log rank test (Graph Pad Prism version 5.03). Virtual images of histology slides were acquired using the scanning program ImageScope (Aperio Technologies, Inc., Vista, CA). Cells were considered positive if the intensity score of the antibody stain was >3. Differences in cytokines, antibody, virus titers, and percentages of positive flow cytometry staining or epithelial regeneration were compared between pregnant and nonpregnant animals by using the two-tailed Student's t test (Excel for Windows 7). Differences in positive immunohistochemistry staining or nitrite concentrations between naïve and virus-infected animals were determined by one-way analysis of variance (ANOVA) followed by Bonferroni's multiple comparison tests (Graph Pad Prism version 5.03). All data are presented as means ± standard deviations.

RESULTS

Increased lethality due to pandemic A (H1N1) 2009 influenza virus is not associated with increased virus replication in pregnant animals.

We determined the severity of pH1N1 infection in pregnant mice (12) using the wt pH1N1strain, A/California/04/09. A 10⁶ EID₅₀ of pH1N1 caused increased mortality in pregnant BALB/c mice compared to nonpregnant controls (P = 0.01) (Fig. 1A). Only 10% of pregnant mice (or mice that delivered pups within 10 days of infection) survived for the duration of the study (26 days), whereas 66% of nonpregnant animals survived. Infections with lower doses of virus showed that the calculated LD₅₀s for pregnant and nonpregnant mice were 10^5.6 EID₅₀ and >10⁶ EID₅₀, respectively. To determine if greater virulence was associated with elevated virus replication in pregnant mice, we measured virus titers. Interestingly, there was no quantifiable difference in lung titers between pregnant and nonpregnant mice at 3 (P = 0.42), 5 (P = 0.78), or 7 (P = 0.12) dpi (Fig. 1B). To confirm these observations, BAL fluid was also tested. There was also no difference in virus titers in BAL fluid between pregnant and nonpregnant animals at 3 dpi (P = 0.09) (data not shown).

Fig. 1.

A/California/04/09-induced lethality and virus burden in pregnant and nonpregnant mice. Pregnant and nonpregnant control mice were infected with 10⁶ EID₅₀ wt pH1N1 virus. Survival, virus titers, and the distribution of influenza virus in the lungs and nasal cavities were determined. (A) Percent survival was measured between pregnant and nonpregnant animals for 26 days post-virus inoculation. *, P = 0.01 for survival comparison between groups of mice. (B) Lung virus titers were assessed by TCID₅₀. P ≥ 0.39 for comparison of virus titers between the pregnant and nonpregnant groups at 3, 5, and 7 dpi (D3, D5, and D7, respectively) (n = 4 or 5). (C) Immunohistochemical staining of influenza virus nucleoprotein (NP) in representative lungs (L) and nasal cavities (N) of infected mice at 3 dpi. Determination of the percentage of positive cell staining was performed by ImageScope (Aperio). Error bars represent standard deviations. Original slide magnifications: lung, ×220; and nasal cavities, ×400. Data are representative of two independent experiments. NS, not significant.

We next asked whether the distributions of virus differed in the nasal cavities and lungs. IHC staining showed characteristic influenza virus NP expression distributed in the epithelial cells lining the lower bronchi, the bronchioles, and randomly throughout the alveolar spaces of pregnant and nonpregnant mouse lungs (Fig. 1C, panels L). The percentages of cells staining positive for NP in the lungs did not differ between pregnant and nonpregnant mice 3 dpi (P = 0.96), as quantified by a computer-based algorithm (ImageScope). These findings were consistent in the nasal cavities, where NP staining in all mice was disbursed exclusively within epithelial cells (P = 0.12) (Fig. 1C, panels N). No difference in levels of NP staining was seen 7 and 9 dpi (data not shown). To detect if pH1N1 spreads systemically, we measured titers of total homogenates from liver, spleen, kidneys, and brain. No infectious virus was detected in any organ other than the lung from any of the mice. Consequently, and in contrast to what has been previously reported (12), our study reveals that increased morbidity due to infection with pH1N1 is not associated with elevated virus replication in pregnant mice.

Lethality in pregnant mice correlates with elevated cytokines, macrophages, and neutrophils.

Severe influenza infection in human and animal models is associated with increased pulmonary proinflammatory cytokine/chemokine expression (7, 12, 28, 40, 62). To assess these immune responses throughout the course of several days, we measured pro- and anti-inflammatory cytokines/chemokines in response to pH1N1 infection. We observed that pH1N1-infected pregnant mice had elevated levels of numerous cytokines in BAL fluid (e.g., macrophage inflammatory protein 2 [MIP-2], MIP-1α, interleukin-6 [IL-6], and tumor necrosis factor alpha [TNF-α]) (P ≤ 0.03) (Fig. 2). Cytokines associated with recruitment of immune cells—macrophage colony-stimulating factor (M-CSF), MIP-1α, and MIP-1β—were also elevated in infected pregnant mice. We observed that pregnant animals developed increased levels of the chemoattractant for monocytes, MCP-1, in the lung (Fig. 2) and blood (P = 0.03) (data not shown). There was no difference in the levels of the cytokines/chemokines between mice at 3 dpi: instead, the difference began appearing at 5 dpi (Fig. 2). Interestingly, although the antiviral cytokine gamma interferon (IFN-γ) appeared elevated 5 and 6 dpi in pregnant mice, there was no statistical difference between groups at any time point tested (P ≥ 0.12). Canonical Th2 cytokines IL-4, IL-5, and IL-6 were increased at 5 or 6 dpi in infected pregnant mice (P ≤ 0.02). Only at a later time point (9 dpi) was the anti-inflammatory cytokine IL-10 elevated in pregnant animals (P = 0.03). The mean value for each cytokine for mock-inoculated animals is indicated as a horizontal dotted line. Figure 2 cytokines did not differ between mock-inoculated pregnant and nonpregnant mice (P ≥ 0.07). What is noteworthy is that the canonical T-helper cell 17 (Th17) cytokine IL-17 and the Th2 cytokine IL-13 were both elevated in the blood of infected pregnant mice (P = 0.008 and P = 0.007) (Fig. 3).

Fig. 2.

Pulmonary cytokine/chemokine expression during acute viral infection. Supernatants from BAL samples were prepared from pregnant (gray circles) or nonpregnant (black circles) mice 3 to 9 dpi, and a panel of 32 cytokines/chemokines was measured. The pg/ml concentration of macrophage colony-stimulating factor (M-CSF), macrophage inflammatory protein 1 alpha (MIP-1 α), macrophage inflammatory protein 1 beta (MIP-1β), macrophage inflammatory protein 2 (MIP-2), monocyte chemoattractant protein 1 (MCP-1), tumor necrosis factor alpha (TNF-α), gamma interferon (IFN-γ), interleukin-4 (IL-4), IL-5, IL-6, keratinocyte-derived chemokine (KC), and IL-10 was determined by Luminex xPonent software. *, P < 0.05 (n = 4 or 5). Error bars represent standard deviations. Dotted lines indicate values from mock-inoculated animals. Data are representative of three independent experiments.

Fig. 3.

Serum cytokine expression during acute infection. Sera were collected from pregnant (white) or nonpregnant (black) mice 3 or 9 dpi, and a panel of 32 cytokines/chemokines was measured. The pg/ml concentration of interleukin-13 (IL-13) and IL-17 was determined by Luminex xPonent software. *, P < 0.05 (n = 4 or 5). Error bars represent standard deviations. Data are representative of three independent experiments.

The severity of 1918 H1N1 and highly pathogenic H5N1 viruses correlates with excessive macrophage and neutrophil pulmonary infiltration (62). In our current studies, the chemoattractants for polymorphonuclear leukocytes (neutrophils) (MIP-1 and keratinocyte-derived chemokine [KC]) or macrophages (MCP-1) were elevated in the lungs of pH1N1-infected pregnant mice (Fig. 2). Therefore, we determined whether these cells were trafficking into the lungs of pregnant animals in greater numbers. First, we observed more interstitium pneumonia in pregnant mice. An influx of inflammatory cell infiltrates was more pronounced in the lungs of infected pregnant mice than that in nonpregnant animals, as seen by H&E staining (data not shown). To determine the identity of these infiltrates, neutrophils, macrophages, and DCs were quantified by flow cytometry from the same lungs that were assessed for cytokines in Fig. 2. As predicted, there was a significant increase in the total number of viable cells in the BAL fluid of pregnant mice at 5 dpi (P = 0.01) and 6 dpi (P = 0.01) (Fig. 4A). Surprisingly, neutrophils were not increased in the BAL fluid of infected pregnant mice at any time point (P ≥ 0.06) (Fig. 4B), and neutrophil levels did not differ between mock-inoculated pregnant versus nonpregnant mice (P = 0.40). Because exacerbated recruitment of neutrophils into the BAL fluid or whole lung is implicated in the severity of influenza infection (16, 62), and there was an increase in the neutrophil recruitment factor KC (Fig. 2), we tested the possibility that neutrophils, although not elevated in BAL fluid of pregnant mice, may be recruited in greater numbers at other sites of the lung tissue. To address this, BAL fluid was collected from lungs of infected pregnant and nonpregnant mice. The same lungs were subsequently stained for neutrophils by IHC. Compared with mock-inoculated animals, all infected mice had more neutrophil foci sporadically disbursed throughout alveolar spaces and interstitium (Fig. 5). The distribution of these cells remained mostly within these areas in all mice. However, the percentage of cells staining positive for neutrophils was higher in the lungs of pregnant mice (P < 0.05), suggesting that these animals have elevated levels of neutrophils in regions outside the bronchial spaces (BAL fluid).

Fig. 4.

Flow cytometric quantification of cells in the lung. Total single-cell suspensions from BAL fluid were collected from pregnant and nonpregnant animals 3 to 9 dpi (n = 4 or 5 per group). Cells were stained with markers for macrophages, neutrophils, and DCs as indicated in Materials and Methods. (A) Total numbers of viable cells in the BAL fluid of infected pregnant (white bars) or nonpregnant (black bars) mice from postinfection days 3 to 9 (D3 to D9). (B) Percentages of neutrophils recruited into the lungs of pregnant versus nonpregnant animals. (C) Macrophage percentages in the lung differ between groups of mice (*, P ≤ 0.001). Dotted lines indicate values from mock-inoculated animals. Error bars represent standard deviations. Data are representative of three independent experiments.

Fig. 5.

Immunohistochemical staining of neutrophils invading the lung. Shown is the staining of representative lung sections isolated from mock-inoculated or infected mice after BAL collection. (A) Ly6g (neutrophil); (B) percentage of cells positive for Ly6g staining between mock-inoculated and infected pregnant and nonpregnant mice. Original slide magnifications, ×220. Each slide is representative of n = 5 mice per group. Data are representative of three independent experiments.

In concordance with our results in Fig. 4A, we observed that macrophages accounted for a greater percentage of total viable cells in the BAL fluid of pregnant mice at 5 and 6 dpi (P ≤ 0.001) (Fig. 4C). In contrast, mock-inoculated pregnant and nonpregnant mice had equivalent percentages of macrophages (P = 0.62). We next confirmed the finding in Fig. 4C by IHC (Fig. 6A). Infected mice were compared to naïve animals. In the lungs of infected pregnant animals, we saw an increase in the number of cells staining positively as macrophages (P < 0.05) (Fig. 6A). Predictably, no difference was seen between naïve groups (P > 0.05). An increase in macrophages by IHC was associated with a considerable amount of macrophages that displayed a round shape, which is characteristic of alternatively activated macrophages (AAMs) (27). These cells were disbursed throughout the lung parenchyma of pregnant animals. To test the hypothesis that these cells were alternatively activated or “type II,” we stained for the type II-associated enzyme arginase-1. This enzyme was distributed around the bronchioles and lung parenchyma (Fig. 6B), thus supporting our hypothesis. This was confirmed quantitatively where the percentage of cells positive for arginase-1 staining was higher in the lungs of pregnant mice than that in nonpregnant animals (P = 0.004). These findings strongly suggest that the increase in lung cytokines/chemokines recruited an excess of macrophages and neutrophils to various pulmonary spaces in pregnant mice.

Fig. 6.

Immunohistochemical staining of macrophages and arginase-1 in the lung. Shown is the staining of representative lung sections isolated from mock-inoculated or infected mice. (A) F4/80 (macrophage); (B) arginase-1 expression between mock-inoculated or infected pregnant and nonpregnant mice. Original slide magnifications, ×220. Each slide is representative of n = 4 or 5 mice per group. Data are representative of two independent experiments.

In agreement with a recent study of severe influenza infection (13), we observed a significant decrease in the percentages of lung DCs at 6 and 7 dpi in severely infected pregnant mice (P ≤ 0.01); however, at 8 dpi, pregnant animals had the greatest percentage of this cell type (Fig. 4D) (P = 0.02). These data imply that an excess of macrophages and neutrophils contributes to elevated pulmonary inflammation in pH1N1-infected pregnant mice.

An increase in pulmonary infiltrates is associated with reduced lung repair in pregnant mice.

Because excess cytokines and increased cellularity contribute to tissue damage in severely infected animals (73, 83), we tested whether pH1N1 infection could elicit enhanced pulmonary injury in pregnant mice. To address this, lung sections were stained with H&E, and the level of epithelial regeneration was analyzed. As predicted, we observed multifocal necrotizing bronchitis and bronchiolitis in both pregnant and nonpregnant infected mice compared to mock-inoculated animals. Epithelial regeneration in the form of flattened and attenuated epithelium (arrows) was evident throughout the majority of lungs from both pregnant and nonpregnant infected animals (Fig. 7A). However, in pregnant mice, the average number of bronchioles denuded of epithelium was greater (P = 0.005). As predicted, no bronchioles in mock-inoculated animals were denuded of epithelial cells. From the total number of bronchioles from pregnant infected animals, 12.5% were denuded of epithelium, whereas, 7.2% of bronchioles of nonpregnant infected mice displayed epithelial loss. These findings indicate pH1N1 infection results in reduced epithelial regeneration in the lungs of pregnant mice.

Fig. 7.

Epithelial repair and nitrite expression in the lungs. (A) The rate of epithelial regeneration of the bronchial regions was determined by H&E staining. Shown are representative lung sections isolated from mock-infected or infected mice. Flattened epithelium is indicated with arrows. Original slide magnification, ×400. Each slide is representative of n = 5 mice per group. An average of the total number of denuded bronchioles from pregnant and nonpregnant mock-infected or infected animals is indicated. (B) Nitrite levels were measured in the BAL fluid at 3, 5, and 7 dpi (D3, D5, and D7, respectively). Data are representative of two independent experiments.

Because we hypothesized that increased recruitment of macrophages and neutrophils plays an important role, we investigated the mechanism by which these cell types may be contributing to severity of infection. A common means that macrophages and neutrophils use to promote severe infection is through injury to lung tissue. The expression of reactive nitrogen species such as nitric oxide (NO) by these cell types plays an important role in lung pathology (19, 21, 30, 35, 41). Therefore, we measured metabolism of NO in the BAL fluid of pregnant and nonpregnant mice. Nitrite, a metabolite and indicator for NO synthesis (6, 51, 61), was elevated at 5 dpi in the BAL fluid of pregnant mice compared to that of nonpregnant mice (P < 0.05) (Fig. 7B). Interestingly, this is the same time point where the number of macrophages was increased in the BAL fluid from pregnant animals (Fig. 4C), suggesting that these cells may be the source of the NO.

Severely infected pregnant mice develop equivalent CD8⁺ T lymphocyte activation and increased numbers of CD4⁺ T regulatory cells.

Because pH1N1 caused alterations in the innate immune response in pregnant mice, we determined whether this had an impact on the adaptive immune responses to infection. We found no differences in the percentages of CD8⁺ T lymphocytes in the spleen, MLN, or BAL fluid between pregnant and nonpregnant mice (P ≥ 0.05) (Fig. 8A, bar graphs). As predicted, the greatest percentage of CD8⁺ T lymphocytes in all infected animals was in the BAL fluid. No quantifiable difference in the percentages of NP tetramer-positive CD8⁺ T lymphocytes was observed between pregnant and nonpregnant animals in any organ tested (P ≥ 0.19) (Fig. 8A, bar graphs and dot plots). We next tested whether the percentages of activation of tetramer-positive CD8⁺ T lymphocytes varied between mice. There was no variation in CD69 expression on tetramer-positive CD8⁺ T lymphocytes from any organ in pregnant compared to nonpregnant mice (Fig. 8B, bar graphs and dot plots) (P ≥ 0.10). These data indicate that magnitude of the CD8⁺ T lymphocyte response to influenza infection did not differ between pregnant and nonpregnant mice at the site of infection nor in the periphery. Elevated numbers of Tregs have been observed in many animal models of pregnancy (2, 69, 70). We observed no difference in percentages of CD4⁺ T lymphocytes in all tissues examined (P ≥ 0.05) (Fig. 9). However, pregnant mice demonstrated a greater percentage of Tregs in the BAL fluid (P = 0.01) but not spleen (P = 0.05) or MLN (P = 0.10) (Fig. 9).

Fig. 8.

CD8⁺ T lymphocyte responses in the lung and lymphoid organs. Shown are the results of flow cytometric analysis of cells in the spleen, MLN, and BAL fluid from pregnant and nonpregnant infected animals 9 dpi (n = 4 or 5). Cells were stained with the appropriate flow cytometric antibodies as described in Materials and Methods. (A) Percentages of total CD8⁺ or CD8⁺ tetramer⁺ T lymphocytes measured in spleen, MLN, and BAL fluid of pregnant versus nonpregnant mice. (B) CD8⁺ tetramer⁺ T lymphocytes that were expressing CD69. Each dot plot figure is representative of 4 or 5 mice per group. Error bars represent standard deviations. Data are representative of two independent experiments. NS, not significant.

Fig. 9.

CD4⁺ T lymphocyte responses in the lung and lymphoid organs. Shown are results of flow cytometric analysis of cells in the spleen, MLN, and BAL fluid from pregnant and nonpregnant infected animals 9 dpi (n = 5). Cells were stained with the appropriate flow cytometric antibodies as described in Materials and Methods. (A) Total CD4⁺ T lymphocytes. (B) Percentages of Tregs were identified with FoxP3 and CD25 antibodies (Treg markers). Significantly higher double-positive staining was observed in the BAL fluid (P = 0.01). Each dot plot figure is representative of 5 mice per group. Error bars represent standard deviations. Data are representative of three independent experiments. NS, not significant.

Pregnant and nonpregnant mice mount equal serum antibody titers to either wild-type pH1N1 or inactivated pH1N1 vaccine.

In pregnant mice, acute immunological differences did not dampen the magnitude of the virus-specific antibody response. Pregnant and nonpregnant mice demonstrated equivalent titers of pH1N1-specific HI, Ig and IgG1, IgG3, IgG2a, IgG2b, and IgG2c in the blood after infection with a sublethal dose of pH1N1 (Fig. 10A, B, and C). Similarly, there was no difference in HI titers (P = 0.10) or total Ig titers (P = 0.07) at 9 dpi in mice infected with a lethal dose of pH1N1 (data not shown). To determine whether the immune response would be altered after vaccination in pregnant animals, female mice primed with the seasonal H1N1 influenza virus strain Brisbane were subsequently vaccinated intramuscularly (or mock vaccinated) with monovalent inactivated pH1N1 vaccine. Brisbane was used to mirror the scenario in nature, where presumably most pregnant women have been exposed to seasonal influenza viruses. These mice were tested for postvaccination antibody responses and weight loss after a lethal challenge with wild-type pH1N1. All pregnant and nonpregnant mice primed with Brisbane and mock vaccinated demonstrated low levels of pH1N1-specific HI titers (≤26). Pregnant and nonpregnant mice demonstrated similar levels of pH1N1-specific HI titers 3 weeks after vaccination (P = 0.30) (Fig. 10D). These results indicate that vaccine induces similar levels of immunogenicity between pregnant and nonpregnant mice.

Fig. 10.

Antibody responses to wild-type pH1N1 or inactivated vaccine in mice. Sera collected from pregnant or nonpregnant mice postinoculation with wt pH1N1 virus or vaccine were assayed for the presence of pH1N1-specific antibodies by HI or ELISA. (A) There were no differences in HI titers (P = 0.26) (A), total Ig (P = 0.51) (B), or IgG subclass (P ≥ 0.11) ELISA titers (C) in response to pH1N1 infection. (D) There was no difference in HI titers between pregnant or nonpregnant animals vaccinated with monovalent inactivated pH1N1 vaccine (P = 0.30). Error bars represent standard deviations. NS, not significant.

DISCUSSION

The focus of the current study was to define why pH1N1 is more lethal during pregnancy. Our findings confirm previous studies (12) that pH1N1 infection is more severe in pregnant mice. Here, we show that enhanced pulmonary neutrophil and macrophage invasion is associated with the severity of pH1N1 infection in pregnant animals. These results are consistent with the finding that these cell types likely contributed to pandemic 1918 H1N1-associated lethality (62). Conversely, evidence suggests that recruitment of neutrophils into the lungs plays a critical role in controlling severe influenza infection (22, 76). The means by which neutrophils contribute to the severity of pH1N1 infection in pregnant mice is unclear. However, our studies suggest that the recruitment of neutrophils into the lungs of infected pregnant mice may be site specific. This mirrors the finding that cytokine expression or cellular recruitment into the BAL fluid is not always an absolute representation of the immune responses occurring throughout the lung (11, 58, 59). Neutrophils were not elevated in the bronchial tubes (BAL fluid) but rather the alveolar spaces and interstitium of pregnant animals. Therefore, the recruitment of cells into the lung may not result in uniform distribution throughout this organ.

Our findings indicate that lung macrophages are also associated with increased lethality in pregnant mice. Studies show that the polarization of macrophages toward an alternative or “nonclassical” phenotype correlates with increased disease severity (23, 52). A shift toward Th2 responses during pregnancy is often accompanied by the development of AAMs (15, 53, 54). Elevation of the enzyme arginase-1 is characteristic of the presence of AAMs and is increased during pregnancy (37, 67). We observed by IHC staining that during pH1N1 infection, pregnant mice displayed elevated pulmonary arginase-1 staining in association with increased infiltrating macrophages. Our finding that two inducers of alternative activation of macrophages, IL-4 and IL-13 (10), are elevated in pregnant mice (Fig. 2) supports this. Therefore, it is plausible that AAMs in the lungs of pregnant mice may contribute to altered innate immune responses to pH1N1 infection; however, this warrants further investigation. Another hallmark we noted in the influenza-infected lungs of pregnant mice was an elevation in nitrite levels. An equivalent number of neutrophils in the BAL fluid (Fig. 4B) between all infected mice implies that macrophages may have contributed to the increased expression of nitrite in the BAL fluid of pregnant animals. Because AAMs do not secrete NO (10), it is unclear whether the majority of macrophages in the lungs of pregnant mice are alternatively activated, while a small percentage are “classically” activated and secreting high levels of NO. Nevertheless, we hypothesize that in the pregnant lung milieu, there may exist a mixed population of macrophages of various phenotypes in response to pH1N1 infection. Macrophages of either classical or alternative activation phenotypes have been observed previously during severe infection (60). Macrophages are likely mediating their deleterious effects on the lungs of pregnant mice through mechanisms including NO production; downstream of this, there may be reduced epithelial regeneration. A major cause of NO-mediated death has been linked to pulmonary injury (55), and NO has been shown to increase vasodilation, thus leading to vascular leakage and increased pulmonary permeability (17, 38, 39). It is unlikely that one single factor was responsible for increased lethality in our pregnancy-influenza model. Instead, the coordinated efforts of neutrophils, macrophages, NO production, and therefore alterations in the repair mechanisms in the lung collectively caused severe pH1N1 infection in pregnant mice.

The repair mechanism of the lung is an ordered process involving the re-epithelialization of damaged tissue. Macrophages, possibly through the action of NO, are associated with reduced regeneration of the lung epithelium. Thus, the impairment of lung repair is a pathway implicated in enhanced severity of pH1N1 in pregnant mice. Lung distress (44, 56, 72) and even pulmonary capillary leak syndrome (5) have been reported in pregnant women with pH1N1-related morbidity. However, the extent of lung damage and repair in pregnant women throughout the course of pH1N1 infection is unknown. Our data suggest that therapies which modulate inflammation and lung repair processes may aid in the reduction of severe pH1N1 infection in pregnant women.

The reduced regeneration of lung epithelia we observed in influenza-infected pregnant mice did not occur due to increased viral replication. At day 7, we noticed a trend toward higher virus replication in pregnant mice (Fig. 1B). However, this trend was not reproducible in separate repeat experiments. Some findings support our hypothesis that increased morbidity may occur independently of elevated virus replication. As a case in point, patients experiencing severe pH1N1-related symptoms did not demonstrate increased virus replication in the lung (3), suggesting that viral load is not an absolute correlate of severe infection in all cases of pH1N1. Furthermore, a lack of association between viral load and severe disease has been observed in other virus models (46, 63). Conversely, it was demonstrated that infection with the wt pH1N1 strain A/Hong Kong/415742/09 resulted in increased virus replication in pregnant mice (12). We cannot account for this apparent discrepancy between previous findings and our current results. The use of different wt pH1N1 isolates—A/California/04/09 (our studies) versus A/Hong Kong/415742/09—may account for the dissimilarity in findings. In summary, our results suggest that severe pH1N1 infection in pregnant mice may depend upon the function of macrophages, neutrophils, NO, and downstream events, which include reduced epithelial regeneration. The implication is that these two cell types contribute to severe infection through lung injury but not by altering virus replication.

Because virus levels in the lungs and nasal cavities of pregnant mice were equivalent to those in nonpregnant animals, there must exist other pathways responsible for the elevated cytokine phenotype and downstream cellular infiltration. In our present study, pH1N1 infection induces elevated proinflammatory cytokines in pregnant mice, which is in agreement with previous findings (12). This may be the result of increased cell activation or changes in the presentation of antigen to cells resulting in amplified secretion of cytokines in the pregnant lung milieu. The increase of the Th2 cytokines IL-4, IL-5, and IL-6 in the lungs and IL-13 in the blood of pregnant mice supports the idea that these animals acquire a Th2-biased host response after pH1N1 infection. Our finding that MCP-1 is elevated in pregnant mice agrees with previous observations that this cytokine is associated with influenza-related morbidity (1, 43).

Dysregulated cytokine expression and increased cellularity in the lung did not alter T lymphocyte or antibody responses in pregnant animals. Our data indicate that the antibody responses (HI, Ig, and IgG subclass titers) are unaltered in pregnant mice administered either a sublethal or lethal dose of wt pH1N1 or inactivated pH1N1 vaccine. Also, we observed normal numbers of CD4⁺ T lymphocytes or the cytokine associated with antibody-dependent T-helper responses, IL-2 (data not shown). Given that antibody titers did not differ between pregnant and nonpregnant mice administered pandemic vaccine (or virus alone), our results demonstrate that innate, not adaptive immune responses are altered in pregnant mice. Our data support the idea that alterations in innate immune responses create a favorable environment for increased susceptibility to a variety of pathogens during pregnancy (33). Our findings demonstrate that pregnancy alters the innate host response, which changes the course of pH1N1 infection independent of enhanced virus replication. In light of our recent data, acute immune responses must be closely considered for proper pandemic planning in pregnant populations.