TNF-α from inflammatory dendritic cells (DCs) regulates lung IL-17A/IL-5 levels and neutrophilia versus eosinophilia during persistent fungal infection

Abstract

Aspergillus fumigatus is commonly associated with allergic bronchopulmonary aspergillosis in patients with severe asthma in which chronic airway neutrophilia predicts a poor outcome. We were able to recapitulate fungus-induced neutrophilic airway inflammation in a mouse model in our efforts to understand the underlying mechanisms. However, neutrophilia occurred in a mouse strain-selective fashion, providing us with an opportunity to perform a comparative study to elucidate the mechanisms involved. Here we show that TNF-α, largely produced by Ly6c⁺CD11b⁺ dendritic cells (DCs), plays a central role in promoting IL-17A from CD4⁺ T cells and collaborating with it to induce airway neutrophilia. Compared with C57BL/6 mice, BALB/c mice displayed significantly more TNF-α–producing DCs and macrophages in the lung. Lung TNF-α levels were drastically reduced in CD11c-DTR BALB/c mice depleted of CD11c+ cells, and TNF-α–producing Ly6c⁺CD11b⁺ cells were abolished in Dectin-1^−/− and MyD88^−/− BALB/c mice. TNF-α deficiency itself blunted accumulation of inflammatory Ly6c⁺CD11b⁺ DCs. Also, lack of TNF-α decreased IL-17A but promoted IL-5 levels, switching inflammation from a neutrophil to eosinophil bias resembling that in C57BL/6 mice. The TNF-α^low DCs in C57BL/6 mice contained more NF-κB p50 homodimers, which are strong repressors of TNF-α transcription. Functionally, collaboration between TNF-α and IL-17A triggered significantly higher levels of the neutrophil chemoattractants keratinocyte cytokine and macrophage inflammatory protein 2 in BALB/c mice. Our study identifies TNF-α as a molecular switch that orchestrates a sequence of events in DCs and CD4 T cells that promote neutrophilic airway inflammation.

Asthma is a heterogeneous inflammatory disease of the airways, and accumulating evidence suggests an association between fungal exposures and disease severity (1). Aspergillus fumigatus, which induces allergic bronchopulmonary aspergillosis (ABPA) in 25% of severe asthmatics, is one of the most common fungi involved (1, 2). Neutrophilic airway inflammation is now well recognized as being a feature of severe asthma (3–5) and ABPA (6, 7). During acute infection, neutrophils are crucial for antifungal defense (8, 9), and lack of neutrophils is the best appreciated risk factor for invasive aspergillosis (10). However, uncontrolled neutrophil infiltration in Aspergillus-induced allergic disorders positively correlates with tissue damage and loss of lung function (6, 11). Because the control of neutrophils is highly desirable in these diseases, there is considerable interest in understanding the mechanisms that promote and perpetuate pulmonary neutrophilia (12).

Individuals inhale A. fumigatus conidia, called resting conidia, into the lung on a daily basis, where these resting conidia germinate into immunogenic swollen conidia (SC) that trigger host immune responses. Under normal circumstances, hosts eradicate them successfully without developing deleterious inflammation because of the highly effective and tightly regulated defense mechanisms. Innate immune cells, mainly macrophages and neutrophils (8, 13), as well as adaptive CD4⁺ Th1 cells (14, 15) mediate antifungal responses. However, both Th2 and Th17 responses are thought to be detrimental, contributing to inflammation (14–17). A. fumigatus can trigger aggressive allergic diseases such as ABPA only in individuals with underlying pulmonary disorders in which pathogen clearance is compromised, resulting in a persistent local and systemic inflammatory response that contributes to the disease state (18).

To better understand the mechanisms underlying Aspergillus-induced neutrophilic airway inflammation, mice were repeatedly exposed to A. fumigatus to mimic the continuous release of fungal antigens in patients who fail to effectively clear the fungus. With this exposure protocol, we observed that two inbred strains of mice, BALB/c and C57BL/6, developed distinct inflammation profiles in the airways, with the former exhibiting neutrophil prominence and the latter an eosinophil bias. We thus probed this strain-associated differential granulocyte influx to better understand the immunological basis for neutrophilic airway infiltration. TNF-α, contributed by inflammatory dendritic cells (DCs) and macrophages, was crucial for promoting IL-17A from CD4⁺ T cells and for collaborating with it to up-regulate expression of the neutrophil chemoattractants keratinocyte-derived chemokine (KC) and macrophage inflammatory protein 2 (MIP-2) as well as airway neutrophilia. TNF-α production and inflammatory DC accumulation were blunted in Dectin-1^−/− and MyD88^−/− mice. C57BL/6 DCs with a muted TNF-α profile harbored more NF-κB p50 homodimers, which are repressors of TNF-α transcription. Therefore, our study highlights a central role for TNF-α in promoting neutrophilic airway inflammation in response to persistent fungal exposure.

Results

Repeated Challenges with A. fumigatus Induce Distinct Granulocyte Patterns in BALB/c and C57BL/6 Mice.

Mice were repeatedly challenged with fungal conidia to mimic persistent local antigenic stimulation in individuals incapable of adequate fungal clearance. Multiple challenges reproduced some critical features of allergic disease observed in persistently exposed individuals that were not present after a single exposure and were true for both BALB/c and C57BL/6 mice (Fig. S1 and SI Results). However, there was a marked strain-specific difference in the type of granulocytes recruited to the airways. Despite the lower level of cellular infiltrate in the airways, the C57BL/6 mice did not demonstrate a proportionately lower number of cells in every compartment. As shown in Fig. 1A, the mice exhibited greater airway eosinophilia as detected in the bronchoalveolar lavage fluid (BALF). In contrast, the BALB/c mice showed a prominent neutrophil-biased response (Fig. 1A and Fig. S1D). Although inflammatory lesions in C57BL/6 mice spared the conducting airways and were present primarily within the parenchyma adjacent to terminal bronchioles, BALB/c mice harbored lesions involving the large and small airways that obliterated extensive portions of the alveolar spaces. BALB/c bronchioles contained degenerate neutrophils admixed with sloughed epithelial cells, fibrin, and cellular debris. The lungs from BALB/c mice had prominent perivascular infiltrates containing dense sheets of neutrophils admixed with smaller numbers of eosinophils. In contrast, the lungs from challenged C57BL/6 mice showed modest multifocal intraalveolar aggregates of macrophages and occasional multinucleated giant cells intermixed with small clusters of degenerate eosinophils and neutrophils (Fig. 1B). This strain-associated difference was only revealed during repeated challenges, with similar neutrophilic influx being observed in both strains soon after administration of 10 × 10⁶ conidia (Fig. S1E). These results suggested that mechanisms that promote neutrophilic versus eosinophilic inflammation come into play during persistent antigenic stimulation. Importantly, the accumulated neutrophils in the BALB/c mice after repeated fungal exposure seemed not to be of any added benefit because the C57BL/6 mice with much less neutrophilia did not show compromised fungal clearance in the long term (Fig. S1F). It is well known that excessive neutrophils can cause indiscriminate damage to the host. For instance, when released into the extracellular fluid by leakage or by cell lysis, neutrophil myeloperoxidase (MPO) is toxic to the tissue (19). Therefore, the significantly higher MPO in the cell-free BALF of BALB/c mice (Fig. 1C) likely contributed to the lung damage shown in Fig. 1B. The inflammatory pattern in BALB/c mice was reminiscent of the neutrophil-rich inflammation in severe asthma and ABPA (6, 20), which persuaded us to investigate the underlying mechanisms.

Fig. 1.

Characterization of the inflammatory responses induced by repeated A. fumigatus challenges in BALB/c and C57BL/6 mice. Mice were sensitized with 10 × 10⁶ conidia, challenged with 1 × 10⁶ conidia eight times, and sampled at 24 h after the last challenge. (A Right) Differential cell counts in the BALF. Eos, eosinophils; Neu, neutrophils; Lym, lymphocytes; Macs, macrophages. (Left) Representative images of cytospin are shown, and the white and black arrows indicate neutrophils and eosinophils, respectively. (B) Representative hematoxylin/eosin staining of lung sections. (Magnification: ×200; scale bars: 50 mm.) (C) MPO activity in the cell-free BALF. Results are represented as mean ± SD (n = 4 mice per group). Shown is an experiment representative of at least two experiments.

IL-17A Contributes to Neutrophil Influx in both Strains.

IL-17A is known to promote neutrophil-rich responses (21). In our model, IL-17A⁺ cells were detectable after the first challenge and expanded during multiple challenges (Fig. 2A). IL-17A is not exclusively produced by CD4 T cells, and a recent study showed that, by 6 d postinoculation with 5 × 10⁶ A. fumigatus conidia, γδ rather than αβ T cells are a crucial source of IL-17A in the lung (17). In our model, at an early time point (24 h after the first challenge), CD4⁺ T and γδ T cells represented ∼40% and ∼13% of IL-17A–producing cells, respectively (Fig. 2A). CD4⁺ T cells became a predominant source of IL-17A after eight challenges, responsible for > 60% of IL-17A⁺ cells, whereas γδ T cells remained ∼10% of IL-17A⁺ cells in both strains (Fig. 2A). Systemic blockade of IL-17A in BALB/c mice by anti–IL-17A antibody or IL-17RA deficiency on the C57BL/6 background selectively reduced neutrophil counts in the BALF by ∼60% (Fig. 2B), suggesting that IL-17A was functional in both strains. We further asked whether differences in IL-17A between the two strains could explain the differential neutrophilia. Because CD4⁺ T cells were the main IL-17A-producers (Fig. 2A), we evaluated Th17 development. Interestingly, the C57BL/6 mice mounted a normal, if not even better, Th17 response, with a slightly higher percentage of Th17 cells within the CD4⁺ T population, a stronger cytokine-secreting capacity (higher mean fluorescence intensity of the positive cells) (Fig. 2C), and equivalent mRNA expression in the CD4⁺ T cells (Fig. 2D). There was a slightly lower IL-17A in the lung homogenate of C57BL/6 mice (Fig. 2E), which might be because of the lower number of lung CD4⁺ T cells in this strain. Thus, IL-17A was important in promoting neutrophilia, regardless of the genetic background. However, the difference in IL-17A production in the two strains did not seem large enough to explain the strain-associated drastic difference in airway neutrophilia. Given that in vitro studies have suggested synergism between IL-17A and TNF-α in the induction of factors that support development and recruitment of neutrophils (22–24), we investigated whether TNF-α contributed to the neutrophil-biased airway inflammation in the BALB/c strain.

Fig. 2.

IL-17A is involved in driving neutrophilia in both strains. (A) Representative FACS plots show intracellular IL-17A staining in the lung cells from PBS control and challenged mice at 24 h after the first and eighth challenges. IL-17A⁺ gates were drawn, and CD4 or γδ T-cell receptor (γδTCR) expression is shown within the IL-17A⁺ cells. The numbers indicate the percentages of cells in the gated populations. (B) BALB/c mice treated with isotype or anti–IL-17A antibody (Ab), IL-17RA^−/− (on the C57BL/6 background), or WT C57BL/6 mice were challenged, and cell differential counts in the BALF were assessed at 24 h after the eighth challenge. (C) Representative FACS plots show intracellular IL-17A staining in the lung CD4⁺ T cells at 24 h after the eighth challenge. The percentage of IL-17A⁺ cells within lung CD4⁺ T cells and mean fluorescence intensity (MFI) of IL-17A⁺ cells are indicated. (D) IL-17A mRNA in lung CD4⁺ T cells at 24 h after the eighth challenge was expressed as fold increase compared with that in lung CD4⁺ T cells of naïve mice. (E) ELISA of IL-17A in the lung homogenates prepared from mice after the eighth challenge or from PBS controls. Results are expressed as mean ± SD (n = 4 mice per group). Shown is an experiment representative of at least two experiments.

TNF-α, Mainly Derived from CD11c⁺ DCs and Macrophages, Is Important in Determining Strain-Specific Pulmonary Neutrophilia.

In response to repeated challenges, TNF-α was appreciably induced in BALB/c mice but only marginally elevated in C57BL/6 mice compared with their corresponding controls (Fig. 3A). The TNF-α difference was much greater than the difference in IL-17A (Fig. 2E), suggesting that TNF-α could be the determinant in strain-specific neutrophilia. BALB/c mice treated with anti–TNF-α antibody (Fig. 3B) or TNF-α KO BALB/c mice (Fig. S2A) showed decreased neutrophil counts. Interestingly, the decrease in neutrophils was accompanied by a tendency toward higher eosinophil counts (Fig. 3B and Fig. S2A), resembling the granulocyte pattern in the airways of the C57BL/6 strain. The IL-17A level in the lung was decreased by TNF-α neutralization, whereas the level of IL-5, which was higher in C57BL/6 than in BALB/c mice, was increased when TNF-α was absent (Fig. 3C), in agreement with the shift from neutrophilia to eosinophilia. This finding suggested that TNF-α acted as a molecular switch to regulate the neutrophil/eosinophil ratio. The drastic strain-associated TNF-α difference required multiple challenges (Fig. S2B), consistent with the finding that distinct granulocyte patterns only occurred after repeated challenges (Fig. 1A and Fig. S1E). Together, these results suggested that strain-specific regulatory mechanisms triggered as a result of persistent stimulation by fungal antigens determined the differential TNF-α outcome, which, in turn, contributed to strain-associated differential neutrophilia.

Fig. 3.

TNF-α is important for determining strain-specific pulmonary neutrophilia. (A) ELISA of TNF-α in the lung homogenates at 24 h after the eighth challenge. (B) Differential cell counts in the BALF from BALB/c mice that were challenged with fungus eight times and treated with isotype or anti–TNF-α antibody. (C) Lung IL-17A and IL-5 mRNA or IL-5 protein after the eighth challenge. The mRNA is expressed as fold increase compared with PBS controls. (D) CD11c-DTR-EGFP Tg mice were challenged with fungus eight times and maintained without DT (−DT) or treated with DT (+DT), and TNF-α in the BALF or lung homogenate was assayed after 24 h. (E) Representative FACS plots show intracellular TNF-α staining in cells from BALB/c mice at 24 h after the eighth challenge. The CD11c⁺ cells were first divided based on their side scatter (SSC) and differential expression of CD11c into gate 1 (CD11c^low SSC^high eosinophils) and gate 2. Cells in gate 2 were further divided into CD11c⁺ low-autofluorescent (AF^low) DCs and CD11c⁺ high-autofluorescent (AF^high) macrophages (Macs). CD11b vs. TNF-α or Ly6C vs. TNF-α expression is shown in different cell types. The numbers indicate the percentages of cells in the gated populations. (F) ELISA of TNF-α in the culture supernatants. Lung DCs or macrophages were sorted at 24 h after the fifth challenge, and 1 × 10⁵ cells per well of each cell type from each strain were cultured in vitro with and without SC. Results are expressed as mean ± SD (n = 4 mice per group). n.s., not significant. Shown is an experiment representative of at least two experiments.

Next, we investigated the cellular source of TNF-α. Administration of diphtheria toxin (DT) to CD11c-DTR-EGFP transgenic (Tg) mice (on the BALB/c background) after challenges depleted >60% of the CD11c⁺ cells (Fig. S2C) by 24 h, which caused a 50–70% reduction of TNF-α levels in both the BALF and the lung homogenate (Fig. 3D), confirming that CD11c⁺ cells were the major source of TNF-α. Consistent with the results described above (Fig. 3 A and D), in response to in vitro SC restimulation, lung cells from BALB/c mice produced a much larger amount of TNF-α than those from C57BL/6 mice, which was mainly derived from CD11c⁺ but not CD11c⁻ cells (Fig. S2D). In challenged BALB/c mice, the CD11c⁺ fraction could be further divided into eosinophils, DCs, and macrophages (Fig. S3 A and B and SI Results), and both DCs and macrophages were able to produce TNF-α (Fig. 3E). In the DC compartment, TNF-α production was largely restricted to CD11b⁺ DCs and preferentially to Ly6C⁺ DCs (Fig. 3E). The expression of Ly6C reflected their monocytic origin, suggesting that these TNF-α⁺ DCs were newly derived from monocytes, which was reminiscent of DCs producing TNF and inducible nitric oxide synthase (iNOS) (called “Tip-DCs”) (25, 26). However, surprisingly, in our model, the TNF-α–producing DCs did not coexpress iNOS, rather TNF-α⁺ DCs and iNOS⁺ DCs were different DC subsets (Fig. S3C), implying that the TNF-α⁺ DC in our model is a distinct cell type.

TNF-α–Producing DCs Contribute the Most to Strain-Associated TNF-α Difference.

Numerically, BALB/c mice had more DCs and macrophages (Fig. S4A). In the BALB/c strain, the frequency of CD11b⁺Ly6C⁺ DCs, the subset that was enriched for TNF-α (Fig. 3D), progressively increased with repeated challenges (Fig. S4B) but, interestingly, C57BL/6 DCs had a much smaller fraction of this subset, a difference that occurred as early as 24 h after the first challenge (Fig. S4B). The phenotypic difference did reflect a functional difference because purified BALB/c DCs isolated after five challenges secreted more than threefold more TNF-α than did C57BL/6 DCs on a per-cell basis, whereas purified macrophages from both strains produced equal amount of TNF-α in response to SC in vitro (Fig. 3F). Thus, macrophages from the two strains possessed comparable TNF-α–secreting capacity on a per-cell basis, although the recovered numbers were different (Fig. S4A). However, comparatively, DCs derived from the two strains not only differed in their numbers but also in their capacity to produce TNF-α, thereby accounting more for the strain-associated TNF-α difference.

Differential Activation of NF-κB and Toll-Like Receptor 2 (TLR2) Pathway Results in DCs with Differential TNF-α–Producing Capacities.

Next, we sought to dissect the signaling pathways responsible for the strain-based neutrophilic response. TLR2, TLR4, and dectin-1 are the most critical pattern recognition receptors (PRRs) for the recognition of A. fumigatus (27–29). In our BALB/c model, the absence of MyD88 or dectin-1, but not TLR4, significantly reduced the infiltration of inflammatory cells, including neutrophils (Fig. S5A), which suggested that, besides dectin-1, TLR2 rather than TLR4 might be more important. We then more closely examined the generation of TNF-α–producing DCs. MyD88 and dectin-1 signaling were essential because lack of either profoundly diminished TNF-α⁺ DCs (Fig. 4A). Furthermore, lack of dectin-1 signaling reduced the abundance of lung DCs, particularly CD11b⁺ Ly6C⁺ DCs, indicating the importance of dectin-1 in the development of this DC type (Fig. 4B).

Fig. 4.

Differential activation of NF-κB and TLR2 results in DCs with differential TNF-α–producing capacities. (A) Representative FACS plots show intracellular TNF-α staining in DCs from WT, MyD88^−/−, or dectin-1^−/− mice at 24 h after the fifth challenge. (B) Representative FACS plots show lung DC abundance and DC subsets in WT BALB/c or dectin-1^−/− mice at 24 h after the eighth challenge. For DC abundance, plots were gated on CD45⁺ cells, and lung DCs are CD45⁺ CD11c⁺ Siglec-F⁻ AF^low cells. The numbers indicate the percentage of DCs within the live cells. (C) Representative FACS plots show Dectin-1 vs. Ly6C, TLR2 vs. Ly6C, and TLR2 vs. Dectin-1 expression on lung DCs at different time points. Lung CD11c⁺ cells were isolated from PBS control or challenged mice at 24 h after the first and fifth challenges using microbeads. Unless otherwise indicated, lung DCs were gated as described in the legend to Fig. 3E, and the numbers in the FACS plots are the percentages of cells in the gated populations. (D) ELISA of TNF-α in the culture supernatant. Macrophages or DCs sorted 24 h after the fifth challenge were cultured without stimulation (−), with curdlan (dectin-1 agonist), or with Pam3Cysk4 (TLR2 agonist). (E) Plots show the ratio of p50/p65 DNA-binding activity in the nuclei of lung DCs or macrophages at 4 h and 24 h after the first challenge. Results shown are mean ± SD (n = 4 mice per group). (F) The binding complex at the κB3 site using nuclear extracts from DCs prepared 24 h after the first challenge or from DCs of naïve mice was examined by EMSA (excess probe was run off the gel). The letters B and C represent complexes formed with nuclear extracts from BALB/c and C57BL/6 mice, respectively. Data shown are representative of at least two independent experiments.

We next questioned whether DCs from the two strains differentially expressed critical PRRs, which subsequently led to differential TNF-α production in DCs and in the lungs. Based on the data generated (Fig. S5A), we focused on dectin-1 and TLR2. Both were expressed on CD11b⁺ Ly6C⁺ DCs (Fig. 4C), the subset that was enriched for TNF-α. Notably, TLR2 was largely colocalized with dectin-1 on the same cells (Fig. 4C). Between the strains, there was no apparent difference in the baseline expression of these molecules. One challenge with fungal conidia only produced a subtle difference between the two strains. However, repeated challenges resulted in modestly higher dectin-1 but substantially higher TLR2 on BALB/c DCs, collectively resulting in a higher frequency of TLR2⁺ Dectin-1⁺ DCs in the BALB/c strain compared with the C57BL/6 strain (Fig. 4C). Although both MyD88 and dectin-1 pathways were required for TNF-α production by DCs, dectin-1 signaling alone was not sufficient to induce functionally different DCs because DCs from both strains after repeated challenges produced similar levels of TNF-α in response to the dectin-1 agonist curdlan (Fig. 4D). With evidence of crosstalk between dectin-1 and TLRs (30), our data suggest that higher levels of TLR2 on BALB/c DCs ultimately cause the remarkable difference in TNF-α output between the strains because BALB/c DCs secreted significantly more TNF-α upon stimulation with the TLR2 agonist Pam3Cysk4 (Fig. 4D). Notably, macrophages from both strains mounted a comparable TNF-α response to both ligands (Fig. 4D), consistent with the observation that macrophages expressed comparable levels of the receptors regardless of the strain.

To further understand the differential TNF-α response in the DCs, we investigated more upstream mechanisms. Given that multiple signaling pathways converge on NF-κB in inflammation (31), we investigated whether there was an inherent difference in NF-κB composition in the nuclei of the DCs in the two strains. We chose to examine early time points (4 h and 24 h) after the first challenge, at which time the expression of PRRs was similar but a modest strain-associated difference in DC-produced TNF-α was observed when DCs were restimulated in vitro (Fig. S5B). A higher p50/p65 DNA-binding activity was detected in the nuclei of C57BL/6 DCs compared with that in BALB/c DCs (Fig. 4E). It is known that the binding of p50 homodimers, as transcriptional repressors, to the κB3 site of murine TNF-α promoter blocks TNF-α transcription in the context of LPS-induced hyporesponsiveness (32, 33). No protein binding to the κB3 site was detected by using EMSA with nuclear extracts prepared from lung DCs of naïve mice (Fig. 4F). However, protein binding was detected with nuclear extracts prepared at 24 h after the first challenge (Fig. 4F). Use of anti-p50 and anti-p65 antibodies showed the binding proteins as containing p50 but not p65 (Fig. 4F). In line with the lower TNF-α profile in C57BL/6 mice, the nuclei of C57BL/6 DCs showed greater p50 homodimer binding activity (Fig. 4F), which might also indirectly cause lower TLR2 expression on C57BL/6 DCs because TNF-α promotes TLR2 expression (34). Interestingly, macrophages from these two strains did not show this NF-κB difference (Fig. 4E), consistent with their comparable TNF-α–producing abilities (Fig. 4D).

TNF-α and IL-17A Collaborate to Induce Neutrophilic Inflammation.

We then examined how IL-17A and TNF-α together influenced neutrophil influx. In vitro studies have shown synergistic effects of IL-17A and TNF-α on lung epithelial cells to induce granulocyte-CSF (G-CSF) for neutrophil mobilization and KC or MIP-2 for their recruitment (22–24). Compared with WT mice, the levels of KC, MIP-2, and, to a lesser extent, G-CSF were lower in TNF-α KO mice (Fig. 5 A and B); however, only KC was reduced by IL-17A neutralization (Fig. 5B). KC and MIP-2, but not G-CSF, levels were also lower in the lungs of C57BL/6 mice compared with those in BALB/c mice (Fig. 5B). These data highlighted the role of both IL-17A and TNF-α in inducing neutrophil chemoattractants. Neither TNF-α nor IL-17A influenced eotaxin levels in the lungs. Interestingly, despite lower eosinophilia, BALB/c mice had a higher eotaxin level (Fig. 5B) but a lower IL-5 level (Fig. 3C).

Fig. 5.

TNF-α and IL-17A collaborate to induce airway neutrophilia. (A) In situ hybridization images show lung KC and MIP-2 in WT or TNF-α^−/− mice. (B) ELISA of KC, G-CSF, MIP-2, and eotaxin in the lung at 24 h after the eighth challenge. (C) The numbers of lung IL-17A⁺, CD4⁻IL-17A⁺, and CD4⁺ IL-17A⁺ cells per mouse were determined by intracellular cytokine staining methods at 24 h after the eighth challenge. (D) Lung TNF-α mRNA in mice at 24 h after the eighth challenge. Results are expressed as fold increase over PBS controls. (E) Representative FACS plots show lung DC abundance, DC subsets, and PRR expression on DCs in PBS control or WT BALB/c and TNF-α^−/− mice at 24 h after the eighth challenge. For DC abundance, plots were gated on CD45⁺ AF^low cells (AF^high macrophages were excluded). Lung DCs are CD45⁺ AF^low CD11c⁺ Siglec-F⁻, and the numbers indicate the percentage of DCs within the live cells. The other plots were gated on lung DCs as described in the legend to Fig. 3E, and the numbers indicate the percentages of cells in the gated populations. Results are expressed as mean ± SD (n = 4 mice per group). n.s., not significant. Shown is an experiment representative of at least two experiments.

To better understand the IL-17A–TNF-α collaboration, we examined how they regulated each other. DC-derived TNF-α is believed to synergize with IL-23 to promote an IL-17A response (35–38). Therefore, the TNF-α–producing DCs in our model, with their preferential and continuous expansion during repeated challenges (Fig. S4), might enhance Th17 responses. The diminished lung IL-17A levels caused by neutralization of TNF-α (Fig. 3C) was attributable to the selective reduction of the number of CD4⁺IL-17A⁺ but not CD4⁻IL-17A⁺ cells (Fig. 5C). In contrast, IL-17A showed no regulatory effect on TNF-α (Fig. 5D). Accordingly, up-regulation of IL-5 in the absence of TNF-α was associated with a higher percentage of lung Th2 cells (Fig. S6A). Blockade of TNF-α in ex vivo lung DC/ T-cell cocultures reduced IFN-γ and IL-17A but promoted IL-5 production (Fig. S6B). Collectively, these results demonstrated the important role of TNF-α in regulating the Th1/Th2/Th17 profile and, in turn, the neutrophil/eosinophil balance. Interestingly, TNF-α was found to regulate not only the level of IL-17A but also that of its own. TNF-α deficiency resulted in a lower percentage of lung DCs, especially CD11b⁺Ly6C⁺ DCs (Fig. 5E), and also caused lower TLR2 and dectin-1 expression on DCs (Fig. 5E). These data demonstrated a role of TNF-α in the promotion of inflammatory lung DCs capable of pathogen recognition and, driving its own production, seemingly functioning as a positive feedback loop to expand its source.

Discussion

We studied the mechanisms of fungus-induced airway neutrophilia by persistent exposure of mice to A. fumigatus (summarized in Fig. S7). As described, specific immunoregulatory mechanisms came into play only during the repeated challenges. Therefore, our study is distinct from previous ones that investigated protective immunity immediately after fungal infection (14–17). TNF-α promoter polymorphisms causing differential TNF-α levels have been associated with the onset and severity of asthma (39). Therefore, our finding of differential TNF-α regulation in mice may have relevance in human disease.

A recent study has identified monocyte-derived inflammatory DCs that participate in fungal phagocytosis and T-cell activation early after A. fumigatus infection (40). Here we show that inflammatory DCs producing TNF-α play an important role in shaping the immune response to persistent fungal exposure. Interestingly, in our model, TNF-α–producing DCs did not coexpress iNOS, which distinguishes them from Tip-DCs (25, 26). TNF-α, derived from DCs and macrophages, also promoted the expansion of the pathogen-sensing and TNF-α–producing inflammatory DCs, which might have been through indirect mechanisms, such as the up-regulation of GM-CSF, the critical cytokine that controls the differentiation of DCs from monocytes (41). Alternatively, TNF-α produced by DCs may act as an autocrine factor to promote cell survival, as was shown in the case of macrophages (42), or to induce DC-recruiting chemokines (43). The induction of dectin-1 on DCs by TNF-α in vivo could be through GM-CSF as well, because GM-CSF directly induces dectin-1 expression on macrophages in vitro (44).

TNF-α acts as a molecular switch in our model to regulate the neutrophil/eosinophil balance. The target of the inhibitory effect of TNF-α on eosinophilia seems to be IL-5. It has been reported that, only in the presence of IL-5, intranasal administration of eotaxin induces eosinophil influx because IL-5 primes eosinophil to be responsive to eotaxin (45). This observation might explain why the BALB/c strain, despite a higher eotaxin level, did not show prominent eosinophilia, which might be because of insufficient lung IL-5 levels.

p50 homodimers in LPS-induced tolerance limit deleterious inflammation (32, 33). The generation of “less inflammatory” C57BL/6 DCs with enriched p50/p50 at the κB3 site could also be a self-control strategy, responding to repeated PRR stimulation to avoid deleterious outcomes. Interestingly, it is cell context–specific and not present in macrophages, which has not been previously appreciated. Our findings provide a framework for the clinical heterogeneity in the response of patients to Aspergillus.

The TNF-α producing DC may act as a double-edged sword, such that tight regulation of this cell type is required to achieve an acceptable balance between host protection and immune-mediated pathology. Our findings, therefore, lend a cautionary note to the potential use of injected TNF-α for DC maturation and trafficking to lymph nodes in the context of DC-mediated cancer vaccines. In specific contexts, the risks may outweigh the benefits. Thus, TNF-α production and neutrophilic inflammation may be causally linked in various disorders, which is not yet adequately appreciated in the literature. Our study suggests that controlling the expansion of TNF-α–producing DCs by targeting TNF-α levels may help ameliorate these disease states.

Materials and Methods

BALB/c, C57BL/6, TNF-α^−/−, TLR4^LPS-d, and CD11c-DTR-EGFP Tg mice on the BALB/c background were purchased from The Jackson Laboratory. MyD88^−/− (46) and dectin-1^−/− (47) mice on the BALB/c background were bred at the animal facility at the University of Pittsburgh, and the IL-17RA^−/− mice (21) were bred at Taconic Farms. All mice were housed under pathogen-free conditions and used between 6 and 8 wk of age. All animal experiments were approved by the Animal Care and Use Committee at the University of Pittsburgh. Information on the animal model and all experimental procedures is provided in SI Materials and Methods.