Compartment-Specific and Sequential Role of MyD88 and CARD9 in Chemokine Induction and Innate Defense during Respiratory Fungal Infection

Aspergillus fumigatus forms ubiquitous airborne conidia that humans inhale on a daily basis. Although respiratory fungal infection activates the adaptor proteins CARD9 and MyD88 via C-type lectin, Toll-like, and interleukin-1 family receptor signals, defining the temporal and spatial pattern of MyD88- and CARD9-coupled signals in immune activation and fungal clearance has been difficult to achieve. Herein, we demonstrate that MyD88 and CARD9 act in two discrete phases and in two cellular compartments to direct chemokine- and neutrophil-dependent host defense. The first phase depends on MyD88 signaling because genetic deletion of MyD88 leads to delayed induction of the neutrophil chemokines CXCL1 and CXCL5, delayed neutrophil lung trafficking, and fatal pulmonary damage at the onset of respiratory fungal infection. MyD88 expression in lung epithelial cells restores rapid chemokine induction and neutrophil recruitment via interleukin-1 receptor signaling. Exogenous CXCL1 administration reverses murine mortality in MyD88-deficient mice. The second phase depends predominately on CARD9 signaling because genetic deletion of CARD9 in radiosensitive hematopoietic cells interrupts CXCL1 and CXCL2 production and lung neutrophil recruitment beyond the initial MyD88-dependent phase. Using a CXCL2 reporter mouse, we show that lung-infiltrating neutrophils represent the major cellular source of CXCL2 during CARD9-dependent recruitment. Although neutrophil-intrinsic MyD88 and CARD9 function are dispensable for neutrophil conidial uptake and killing in the lung, global deletion of both adaptor proteins triggers rapidly progressive invasive disease when mice are challenged with an inoculum that is sub-lethal for single adapter protein knockout mice. Our findings demonstrate that distinct signal transduction pathways in the respiratory epithelium and hematopoietic compartment partially overlap to ensure optimal chemokine induction, neutrophil recruitment, and fungal clearance within the respiratory tract.


Introduction
Aspergillus fumigatus is a leading cause of infectious morbidity and mortality in patients with hematologic malignancies, allogeneic stem cell transplant recipients and in patients receiving immune suppressive therapies [1,2]. Airborne conidia (asexual spores) represent the infectious propagules that humans inhale daily. Due to their small size (*2-3 μm diameter), conidia can bypass mucociliary clearance and reach terminal airways [3,4]. In immune competent individuals, lifelong asymptomatic clearance of inhaled conidia prevents the formation of tissueinvasive hyphae.
Neutrophils represent essential effector phagocytes that are rapidly recruited to infected airways and act against A.fumigatus conidia and hyphae [5][6][7]. Neutrophil recruitment to the portal of infection depends on rapid chemokine induction, with a prominent role for ELR + CXC chemokines and their receptors in this process. In mice, antibody-mediated blockade [8] or genetic deficiency of the chemokine receptor CXCR2 [7] results in delayed neutrophil airway recruitment and the development of invasive aspergillosis (IA). Consistent with this notion, overexpression of a CXCR2 ligand (i.e. CXCL1/KC) in the murine lung improves outcomes in neutropenic mice infected with A. fumigatus [9].
IL-1R and most TLR superfamily members activate the signal transducer MyD88 (Myeloid differentiation primary response gene 88). Similar to CARD9, MyD88 activation following A. fumigatus challenge has been linked to the induction of NF-κB-dependent production of CXCR2 ligands in vitro and in vivo [10,23,[27][28][29]. Although TLR2-and TLR4-deficient macrophages and dendritic cells show defects in the production of CXCR2 ligands in vitro [10,11,30], the relative contribution of TLR-MyD88 versus IL1R-MyD88 signaling in neutrophil recruitment and in protective innate immunity in the lung following respiratory A. fumigatus challenge remains to be elucidated.
During respiratory fungal challenge, the mechanism by which the host integrates signals from CLR/CARD9, IL-1R1/MyD88, and TLR/MyD88 pathways to achieve rapid chemokinedependent influx of effector phagocytes and conidial clearance remains undefined. One possible scenario is that CARD9-and MyD88-dependent signals act within the same cellular compartment and in the same time frame for optimal chemokine induction, as has been demonstrated for Dectin-1/CARD9 and TLR2/MyD88-dependent macrophage TNF and IL-12 release in response to the model fungal particle zymosan [21,31]. In a second scenario, MyD88-and CARD9-coupled signals may act in distinct cellular compartments in the lung within the same time phase post-infection (p.i.) to induce chemokines and coordinate innate immune responses. The latter scenario is exemplified by Pseudomonas aeruginosaor Legionella pneumophila-induced IL-1α/IL-1β production by hematopoietic cells; this process, in turn, rapidly activates IL-1R1 signaling in pulmonary epithelial cells to promote airway neutrophil influx [32][33][34]. However, since gram-negative bacteria associated with pneumonia are not known to induce CLR/CARD9 signaling in vivo, conclusions derived from the above studies may not extrapolate to respiratory fungal infection models. A common prediction of the first and second scenario is that a defect in either signaling pathway would diminish chemokine induction and neutrophil recruitment in the same time period p.i. These scenarios could be distinguished by identifying common or distinct cellular compartments in which the signaling pathways operate.
Alternatively, CLR/CARD9 and TLR/MyD88 or IL1R1/MyD88 signaling pathways may act primarily in temporally distinct phases, either in the same or in distinct cellular compartments, creating two additional scenarios. In these scenarios, each protein would mediate a distinct phase of chemokine induction and neutrophil recruitment that is largely independent of the phase induced by the other signaling adapter. For example, MyD88-and type I interferon signaling induce sequential phases of CCL2/MCP-1, a chemokine that regulates inflammatory monocyte trafficking during systemic listeriosis [35].
In this study, we demonstrate that IL1R1/MyD88 signaling plays a crucial role in innate host defense by mediating rapid CXCL chemokine induction and airway neutrophil recruitment during respiratory fungal infection. This initial phase of chemokine induction and effector cell recruitment occurs independently of CARD9 signaling and depends on IL-1R and MyD88 expression within lung epithelial cells. The mortality defect observed in MyD88 (−/−) mice can be significantly ameliorated by administration of a single dose of recombinant CXCL1 at the onset of infection, consistent with the notion that MyD88-coupled signals act to orchestrate neutrophil recruitment. In contrast, CARD9 signaling within hematopoietic cells mediates CXCL chemokine induction and neutrophil recruitment in the second phase, with neutrophils representing a major cellular source of CXCL2. Since cell-intrinsic MyD88 and CARD9 expression in neutrophils are dispensable for the induction of neutrophil conidiacidal activity, our data indicate that MyD88-and CARD9-coupled signals act primarily to orchestrate sequential phases of neutrophil recruitment during respiratory A. fumigatus infection. Thus, our data illustrate an essential role for lung epithelial cells in innate antifungal immunity and delineate compartment-specific and additive roles for MyD88 and CARD9 signaling that collectively regulate chemokine induction and neutrophil trafficking during pulmonary infection with A. fumigatus.
Histopathologic analysis demonstrated differences in the pattern of inflammation between the two groups at 48 h p.i. (S1A-S1B Fig. in S1 Text). Within WT lung sections moderate to severe, multifocal to coalescing inflammation was widespread (S1A and S1C Fig. in S1 Text). Although inflammatory foci in MyD88 (−/−) lung sections were not as widespread as in control mice, the foci were admixed with cellular debris indicative of severe tissue necrosis (S1B and S1D Fig. in S1 Text), consistent with the development of necrotizing fungal bronchopneumonia. Although MyD88 (−/−) mice exhibited a higher fungal burden in the lung than WT mice, overt hyphal tissue invasion was not apparent in either group. In sum, these results indicate that early MyD88 function is essential for effective fungal clearance and prevents tissue damage that compromises murine survival.
Using the same experimental set-up, we next compared conidial uptake and killing by MyD88 (−/−) and MyD88 (+/+) neutrophils within the same lung, eliminating differences in cell trafficking and in inflammation observed in MyD88 (−/−) and in C57BL/6 mice. Conidial uptake by and conidial viability in lung neutrophils was statistically indistinguishable in MyD88 (−/−) and MyD88 (+/+) counterparts isolated from BM chimeric mice (S4A-S4B Fig. in S1 Text). Furthermore, in vitro conidial uptake by bone marrow neutrophils (BMNF) isolated from WT and MyD88 (−/−) mice was similar (S4C Fig. in S1 Text). These data indicate that the induction of neutrophil phagocytic and conidiacidal activities is not coupled to neutrophil-intrinsic MyD88 function during respiratory fungal infection.

MyD88 is required for chemokine induction during A. fumigatus infection
To investigate a link between the MyD88-dependent neutrophil recruitment defect and chemokine induction in the lung, we focused on the transcription and translation of the neutrophilrecruiting chemokines CXCL1/KC, CXCL2/MIP-2, and CXCL5/LIX since these chemoattractants act as agonists for the chemokine receptor CXCR2. Administration of anti-CXCR2 antibodies exacerbates mortality [8] and genetic deficiency in CXCR2 leads to delayed neutrophil recruitment following A. fumigatus challenge [7]. Because CXCL1 mRNA stability is regulated in part by inflammatory stimuli [36], we carried out in situ mRNA hybridization on lung sections of 3 × 10 7 Af293 or PBS administered MyD88 (−/−) and WT mice, with gene-specific riboprobes for cxcl1 and cxcl2. Representative high power micrographs from lung sections demonstrated a severely attenuated signal for cxcl1 but not for cxcl2 in infected MyD88 (−/−) mice compared to infected WT mice (Fig. 3A). No riboprobe signal was observed in uninfected controls (S5 Fig. in S1 Text). Although we could not unambiguously identify the cell types that produce CXCL1 and CXCL2 mRNA in infected mice, these data indicate that MyD88-dependent signals are required for CXCL1 but not for CXCL2 mRNA induction or stability in the lung following A. fumigatus challenge.
We next examined cytokine induction in A. fumigatus-infected WT and MyD88 (−/−) at the protein level by ELISA. The lung and BALF tissues in MyD88 (−/−) mice displayed a marked decrease in CXCL1 and CXCL5 levels, but not in CXCL2 levels at 10 h p.i. (Fig. 3B-3C, S6A-S6F Fig. in S1 Text) consistent with a prior study [29]. BALF and lung TNF were increased in MyD88 (−/−) mice. However, a number of other cytokines assayed (IL-12p70, IL-6, IL-17A, IL-1β) were not significantly reduced or elevated in BALF and lung tissues of MyD88 (−/−) mice at 10 h p.i. (Fig. 3B-3C). Thus, consistent with the MyD88-dependent defect in neutrophil recruitment, we observed a MyD88-dependent reduction of the chemokines CXCL1 and CXCL5 in the lung.
To examine the contribution of lung epithelial cells to chemokine induction and neutrophil recruitment, we analyzed MyD88 (−/−) mice that express MyD88 under control of the club cell 10 kDa protein (CC10) promoter, restricting MyD88 expression to lung epithelial cells [33,37]. CC10 promoter-driven MyD88 expression rescued neutrophil recruitment (Fig. 5E) and CXCL1 and CXCL5 induction (Fig. 5F-5H) in A. fumigatus-challenged mice compared to nontransgenic, globally MyD88-deficient littermate controls. These data indicate that early IL-1R/ MyD88-dependent signals in lung epithelial cells recruit neutrophils to the infection site. To determine whether ELR + CXC chemokine induction and neutrophil recruitment represents the major mechanism by which MyD88 mediates rapid host defense against A. fumigatus conidia, we administered a single dose of recombinant CXCL1 (rCXCL1), CXCL2 (rCXCL2), CXCL5 (rCXCL5), or PBS diluent via the i.t. route to MyD88 (−/−) mice 4 h p.i. and monitored murine survival and neutrophil recruitment. The median survival of rCXCL1-treated mice was 8 days compared to 3 days for PBS-treated controls (Fig. 5I). The prolonged survival in rCXCL1-treated MyD88 (−/−) mice correlated with partial restoration of early neutrophil recruitment (S9 Fig. in S1 Text). In contrast, rCXCL2 or rCXCL5 did not prolong murine survival (S10 Fig. in S1 Text). Thus, murine mortality caused by a global deficiency in MyD88 signaling is partially reversed by administration of a single chemokine at a single time point, illustrating a central role of airway epithelial MyD88-dependent chemokine induction and neutrophil recruitment at the earliest stage of respiratory A. fumigatus infection.

Hematopoietic CARD9 signaling drives a second phase of neutrophil recruitment
In a previous study, we found that CARD9 was dispensable for neutrophil recruitment and CXCL chemokine induction in the first 12 h p.i., However, our group and others found that Dectin-1 (−/−) and CARD9 (−/−) mice display defective chemokine production and neutrophil recruitment at 24 and 36 h p.i., respectively [14,23], time points that lie outside the MyD88-dependent period. We therefore investigated whether MyD88 and CARD9 act in the same or in  (Fig. 6A). Consistent with this finding, BALF CXCL1, CXCL2, and CXCL5 chemokine levels were reduced in CARD9 (−/−) ! WT mice, but not in WT ! CARD9 (−/−) mice (Fig. 6B), indicating that CARD9 acts in hematopoietic cells to mediate the second phase of ELR + CXC chemokine induction and lung neutrophil recruitment. Thus, MyD88 and CARD9 act in different cellular compartments to regulate these processes.
Neutrophils represent the principal cellular source of CXCL2 during the CARD9-dependent phase of recruitment To determine whether fungal stimulation directly induces ELR + CXC chemokine induction in hematopoietic cells, we cultured WT and CARD9 (−/−) BM macrophages (BMMs) with A. fumigatus germlings. BMMs produced CXCL1 and CXCL2 in a CARD9-dependent manner (Fig. 7A), though CXCL5 was not detected in these co-culture experiments.
To visualize the cellular source of CXCL2 in the murine lung, we generated a CXCL2 reporter mouse in which GFP expression was placed under control of the murine CXCL2 promoter (Fig. 7B). GFP transgene expression in BMMs prepared from CXCL2 reporter mice correlated with CXCL2 release triggered by TLR2 agonist stimulation (Fig. 7C). To define the cellular sources of CXCL2 and to examine CARD9-dependent regulation of CXCL2, we crossed to the transgene to the CARD9 (−/−) background and compared its expression in CARD9 (+/+) and CARD9 (−/−) mice infected with A. fumigatus. GFP expression was observed primarily in neutrophils, and to a lower frequency, in inflammatory monocytes and monocyte-derived CD11b + DCs (Fig. 7D). In contrast, GFP-expressing CD3 + , CD19 + , or NK1.1 + cells were not observed in lungs of A. fumigatus-infected mice at 36 h p.i. Lung-infiltrating neutrophils thus represent the most abundant cellular source of CXCL2 and its expression was dependent on CARD9.
To examine the cumulative effect of MyD88 and CARD9 function in host defense, we challenged double knockout, single knockout, and control mice with a conidial dose (3 × 10 7 ) that is sub-lethal for single adapter protein knockout strains. MyD88 (−/−) CARD9 (−/−) mice rapidly and universally succumbed to this infectious inoculum, while the mice in all other groups survived the infection (Fig. 8C). Histopathological examination of hematoxylin and eosin (H&E) stained lung sections from MyD88 (−/−) CARD9 (−/−) mice demonstrated severe disruption of pulmonary architecture and multifocal to coalescing areas of necrosis centered on and around vessels and bronchioles (Fig. 8Di-8Dii). Necrotic areas were characterized by abundant cellular and nuclear debris admixed with many fungal hyphae (Fig. 8Dii, arrow). GAS stained lung sections of MyD88 (−/−) CARD9 (−/−) mice contained many germinating conidia and fungal hyphae expanding and surrounding vessels and bronchioles, indicative of severe tissue destruction associated with IA (Fig. 8Diii). Mice in other groups however, did not develop tissue-invasive hyphae during the course of study (S12 Fig. in S1 Text, arrows show regions of hyphal proliferation). Collectively, these data indicate a sequential and additive role for MyD88 and CARD9 signaling in host defense against A. fumigatus, with critical contributions from lung epithelial and hematopoietic cells in orchestrating chemokine and neutrophil recruitment to prevent tissue-invasive disease.

Discussion
Rapid neutrophil influx is essential to prevent the germination and tissue invasion of airborne conidia that are inhaled daily during human life. A. fumigatus activates multiple innate immune signaling pathways that act in distinct cell types to induce an abundance of neutrophil chemotactic molecules. Deciphering the timing and importance of each component to the orchestration of effective innate immune defense has been challenging. In this study, we demonstrate that IL-1R/ MyD88-mediated signals in lung epithelial cells and CARD9-mediated signals in hematopoietic cells provide sequential contributions to ELR + CXC chemokine induction and to airway neutrophil recruitment. In the absence of either signaling pathway within the relevant cellular compartment, lung ELR + CXC chemokine expression levels are diminished and neutrophil recruitment is either delayed at onset of fungal challenge (i.e. in MyD88 deficiency) or lags in the ensuing phase (i.e. in CARD9 deficiency). Among the ELR + CXC chemokines, loss of MyD88 signaling was associated with defects in CXCL1 and to a lesser extent in CXCL5, while loss of CARD9 signaling was associated with defects in CXCL1 and CXCL2. Although this study and our previous work [23] demonstrate that loss in either signal adapter protein increases murine susceptibility to A. fumigatus challenge, host defense against inhaled conidia was profoundly compromised when both MyD88-and CARD9-mediated signals were absent. Neutrophil-intrinsic MyD88 and CARD9 function is dispensable for the induction of neutrophil phagocytic and conidiacidal activities, supporting the view that the primary role of both adaptors is to orchestrate neutrophil recruitment. This idea is reinforced by the observation that a single application of exogenous rCXCL1 prolongs survival in infected MyD88 (−/−) mice.
Given the essential requirement for CXCR2 ligands in host defense, our study did not focus on other neutrophil chemoattractants that can emerge in the vicinity of microorganisms, e.g. complement component C5a, or at endothelial sites, e.g. leukotriene B4 [39,40]. Other chemoattractants implicated in neutrophil chemotaxis include CCL3, CCL6, and CCL9 (via CCR1), platelet-activating factor (via 2 receptors), and CXCL12 (via CXCR4) [40]. Thus, it is possible that rCXCL1 administration enhances neutrophil recruitment by replenishing endogenous CXCL1 or by compensating for MyD88-dependent deficits in other chemoattractants that may act in series or parallel [39]. However, under the experimental conditions tested, rCXCL2 and rCXCL5 did not ameliorate MyD88-dependent mortality. One possible explanation is that endogenous CXCL2 and CXCL5 were relatively preserved in infected MyD88 (−/−) mice at 10 h p.i., unlike CXCL1, blunting the impact of exogenous administration. Jeyaseelan and colleagues reported that CXCL1 regulates expression of CXCL2 and CXCL5 during murine Klebsiella pneumoniae pneumonia [41]. During murine pulmonary A. fumigatus infection, the early defect in CXCL1, and to a lesser extent CXCL5, in MyD88 (−/−) mice was not associated with reduced CXCL2 levels, consistent with the notion that CXCL2 production in this model is not strictly dependent on CXCL1.
Our results indicate that lung epithelial cells play a central role in coordinating innate immune responses to inhaled conidia. We provide evidence for this statement in the form of experiments that demonstrate a critical role for MyD88 signaling in radioresistant cells, but not in radiosensitive cells, for CXCL1 and CXCL5 induction and neutrophil recruitment. In addition, expression of MyD88 under control of the well-characterized CC10 promoter [33] is sufficient to rescue CXCL1 and CXCL5 induction and neutrophil recruitment. A previous study on inflammatory responses to conidia in immortalized lung cell lines (BEAS-2B) implicated phosphatidylinositol-3-kinase, p38 MAP kinase, and ERK1/2 signaling pathways in the induction of interleukin-8 (IL-8) [42], a human equivalent of murine CXCL1. In that study, transfection of a dominant negative MyD88 mutant gene did not interfere with fungus-induced IL-8 release, though the findings were not extended to primary cells or to the lung tissue context. It is unclear whether co-culture of BEAS-2B cells and A. fumigatus induces IL-1R ligands in vitro. This step however, appears to be critical for epithelial cell activation in the lung and ensuing neutrophil airway trafficking following respiratory A. fumigatus challenge. This view is further supported by neutrophil recruitment studies in the lung following challenge with common bacterial agents of pneumonia including Legionella pneumophila [32,34], Pseudomonas aeruginosa [33], and Streptococcus pneumonia [43]. Additionally, in a cutaneous Staphylococcus aureus infection model, IL1R-MyD88 signaling mediates neutrophil recruitment in a TLR-independent manner [44,45].
How does A. fumigatus trigger IL-1R signaling in lung epithelial cells? Within 2 hours following intratracheal (i.t.) A. fumigatus challenge, alveolar macrophages transcribe Il1α and Il1β genes [46], and bioactive IL-1α and IL-1β is detectable in the lungs at 6 hours p.i. (Caffrey A. et al., manuscript submitted) coincident with the onset of a neutrophilic cellular influx [7]. In a parallel study, Obar and colleagues unveil a dominant role for IL-1α in early lung CXCL1 chemokine induction and neutrophil recruitment (Caffrey A. et al., manuscript submitted). Lung-resident myeloid cells represent a potential source of IL-1α and IL-1β at the earliest time point post-infection. A. fumigatus infection activates the transcription factor hypoxia-inducible factor-1α (HIF1α) in myeloid cells and this process is linked to defective IL-1α and CXCL1 induction, impaired neutrophil recruitment, and murine susceptibility to IA [47]. Epithelial cells may contribute to early IL-1α release as well, particularly in the context of tissue damage or injury, since pro-IL-1α constitutively expressed in lung epithelial cells, and upon necroptotic cell death, is processed via calpain-dependent cleavage and released [48,49]. IL-1α can rapidly induce chemotactic mediators and neutrophil recruitment to sites of inflammation, as demonstrated in sterile models of ischemic and hypoxic injury [48,49]. This model of IL-1α function is expanding to include microbial agents of inflammation, e.g. L. pneumophila within the respiratory tract [34]. Since respiratory fungal infection induces hypoxic tissue environments in the lung [50], receptor-interacting protein kinase 3 (RIP3)-dependent induction of necroptotic cell death [51] by inhaled conidia may provide a stimulus for IL-1α release. Although the impact of A. fumigatus conidia on necroptotic cell death-which favors IL-1α release-has not been investigated in detail, conidial melanin appears to inhibit apoptotic cell death-which favors IL-1β release-in macrophages and in epithelial cells in vitro [52][53][54][55]. Thus, the presence of conidial melanin and modulation of host cell death pathways may shape early IL-1α/β responses during respiratory fungal infection.
MyD88 acts as a transducer not only for IL-1R-mediated signals, but also for other IL-1R family members [48]. Previous studies indicate that A. fumigatus induces the synthesis of IL-1 receptor antagonist (IL-1RA) and the transcription of IL-36γ and IL-36 receptor antagonist in human peripheral blood mononuclear cells [56,57]. Although our studies and those of Obar and colleagues demonstrate a critical role for IL-1R/MyD88 signaling in host defense against A. fumigatus, our findings do not exclude a role for other IL-1R family members, e.g. IL-36R, in linking IL-1R-dependent signals to ELR + CXC chemokine induction.
Our data did not reveal a requirement for CARD9 signaling in IL-1β and CXCL1, CXCL2, and CXCL5 induction and in airway neutrophil recruitment within the first 10 h p.i. In contrast, we and other investigators reported Dectin-1-and CARD9-dependent defects in IL-1α/β and CXCL1 and CXCL2 production at later time points, starting at 24 h post-infection [14,23]. In response to fungal stimuli, Dectin-1/Syk/CARD9 signaling is associated with activation of canonical NLRP3/ASC/caspase-1 inflammasomes [13,24] and of non-canonical ASC/caspase-8-containing inflammasomes [25]; both mediate the conversion of pro-IL-1β to biologically active IL-1β. Obar and colleagues observe that genetic deletion of ASC, critical for assembly of caspase-1-and caspase-8-containing inflammasomes does not diminish lung neutrophil recruitment nor increase susceptibility to IA, in contrast to IL-1R deficiency (Caffrey A. et al., manuscript submitted). Thus, CARD9-dependent production of IL-1β does not appear to be the primary trigger for IL-1R signaling, CXCL1 induction, and neutrophil recruitment in the initial phase of infection, though CLR/CARD9-dependent signals play a significant role in amplifying and sustaining neutrophil recruitment at later time points.
In the respiratory fungal infection model, MyD88-dependent CXCL1 and CXCL5 release correlates with early phase lung and airway neutrophil recruitment. Remarkably, administration of recombinant CXCL1 ameliorates murine mortality in MyD88 (−/−) mice, suggesting that MyD88-dependent control over neutrophil chemotaxis represents the major physiologically relevant function of early IL-1R/MyD88 signaling. In addition, MyD88 signaling negatively regulates TNF induction in A. fumigatus-infected mice, as demonstrated by our study and others [29]. Thus, TNF-induced lung tissue damage and necrosis, associated with defective ELR+ CXC chemokine induction, defective neutrophil recruitment, and defective fungal clearance, likely contributes to observed infectious outcomes.
In our study, CARD9 regulates CXCL1 and CXCL2 release in the second phase of neutrophil recruitment, with neutrophils acting as the primary hematopoietic source of CXCL2. Several explanations are possible for the differential temporal regulation and roles of CXCR2 ligands during pulmonary aspergillosis. First, lung and airway CXCL1 levels increase more rapidly than CXCL2 levels at the onset respiratory fungal infection [23,29] and thus CXCR2-dependent trafficking steps may reflect the relative abundance of cognate ligands. Second, since different resident and recruited cell types (e.g. neutrophils for CXCL2) may represent the major sources of these chemoattractants, the arrival and position of cellular CXCL1 or CXCL2 producers during neutrophil lung influx may determine the relative timing and role of CXCL1 and CXCL2 during this process. In addition, the relative access of individual ELR + chemokines to glycosaminoglycans during fungal challenge is likely to play a key role in forming gradients and directing neutrophil trafficking into the lung [58,59]. To address these issues, the development of a complete set of transgenic reporters and conditional ablation strategies for CXCR2 ligands will facilitate more incisive experiments to determine the relative contribution, anatomic localization, cellular sources, and regulation of individual CXCR2 ligands during respiratory fungal infection.
In humans, Mendelian defects in MyD88 and CARD9 signaling do not lead to the spontaneous development of aspergillosis, unlike individuals with Mendelian defects in NADPH oxidase activity (i.e. chronic granulomatous disease). Children that lack MyD88 signaling develop pyogenic infections, primarily due to Streptococcus pneumonia [60], and individuals with CARD9 deficiency develop mucocutaneous and disseminated candidiasis and dermatophytosis [38,61,62]. These observations support the model that IL-1R/MyD88 and CLR/CARD9 signaling provide redundancy in orchestrating sterilizing immunity against inhaled conidia. Since humans typically inhale several hundred to several thousand conidia daily [63], the presence of a single functional adapter pathway is sufficient to mediate fungal clearance in an otherwise immune competent host. Consistent with this idea, we observe that genetic deletion of both adapter proteins significantly increases murine susceptibility to invasive disease compared to deletion of a single adapter protein. In the context of allogeneic hematopoietic cell transplantation (HCT), genetic variations in the IL-1 gene cluster are associated with susceptibility to IA, though these data must be interpreted cautiously since an independent validation of this cohort has not been reported [64]. Allelic differences in components of CLR/CARD9 and in MyD88 signaling pathways have been linked to susceptibility to IA in allogeneic HCT patients in multiple cohorts [12,[65][66][67][68]. Thus, genetic variants in CARD9-or in MyD88-dependent signaling pathways may predispose to IA only in the setting of innate immune damage (e.g. in patients with neutropenia or in patients that receive corticosteroids).
In sum, our study reveals coordinated regulation of chemokine induction and neutrophil recruitment by IL-1R/MyD88 and CLR/CARD9 signaling pathways that operate in a biphasic manner and in epithelial and hematopoietic compartments to orchestrate sterilizing immunity against A. fumigatus. Further insight into the signals and temporal sequence of events required for innate immune orchestration is likely to inform strategies to identify and intervene in patients at high risk for aspergillosis.

Ethics statement
All animal experiments were conducted with sex-and age-matched mice and performed with approval from the FHCRC (protocol number 1813) and MSKCC (protocol number 13-07-008) Institutional Animal Care and Use Committee. Animal studies were compliant with all applicable provisions established by the Animal Welfare Act and the Public Health Services Policy on the Humane Care and Use of Laboratory Animals.

Reagents and antibodies
Chemicals and cell culture reagents were purchased from Sigma-Aldrich and Gibco respectively. A. fumigatus strain Af293 was used for all experiments. Generation of CXCL2 reporter mice CXCL2 reporter mice were generated by bacterial artificial chromosome (BAC)-mediated transgenesis using the recombineering strategy developed by Heintz and colleagues [70]. The endogenous CXCL2 locus on BAC clone RP23-87C13 (BACPAC Resources Center, Children's Hospital Oakland research Institute) was modified 3' to the terminal Asn residue in the fourth exon to encode the following insertion in the 5' to 3' direction: an HA peptide (-YPYDVPDY-), a -GSG-linker, a 2A peptide (-GSGAPVKQTLNFDLLKLAGDVESNPGP-) and amino acids 1-238 of enhanced GFP followed by a stop codon. The BAC modification procedures were performed as described in [71,72] and the modified BAC was analyzed by DNA sequencing and Southern blotting to verify GFP integration at the CXCL2 locus. The purified modified RP23-87C13 BAC was injected into fertilized B6/129 oocytes at the University of Washington Transgenic Resources Program. Four potential founder animals were identified among offspring screened by PCR. One candidate, designated 3090, transmitted the CXCL2-GFP transgene to one-half of the progeny and was the founder of the CXCL2 reporter colony. CXCL2 reporter mice were backcrossed 10 generations to C57BL/6 mice. To generate CXCL2 reporter mice on the CARD9 (−/−) background, N8 CXCL2 reporter mice were crossed to CARD9 (−/−) mice and littermate non-transgenic CARD9 (−/−) controls were used in experiments.
Perfused murine lungs were homogenized in 2 mL of PBS, 0.025% Tween-20 for colonyforming units (CFUs) and for ELISA. For histology, perfused lungs were fixed in 10% neutralbuffered formalin, embedded in paraffin, sectioned at 4μm and stained with hematoxylin & eosin (H&E) or Gomori's ammoniacal silver (GAS). Slides were reviewed by a board certified pathologist (SEK). Images were captured from whole slide images, acquired with the Aperio ScanScope (Aperio Technologies) using ×20 and ×40 objectives at the FHCRC's Experimental Histopathology Shared Resource. Images for S12 Fig To assess fungal burden, infected mice were euthanized and lungs were immediately frozen in liquid nitrogen. Samples were freeze-dried, homogenized with glass beads on a Mini Beadbeater (BioSpec Products, Inc., Bartlesville, OK), and DNA extracted with E.N.Z.A fungal DNA kit (Omega Bio-Tek, Norcross, GA) or phenol chloroform extraction. RT-PCR was performed as described previously [73]. LDH (CytoTox96 non-radioactive cytotoxicity assay kit; Promega, Cat. No. G1780) and albumin assays (Albumin (BCG) reagent set; Eagle Diagnostics, Cat. No. 2050-1) were performed on BALF collected from infected mice using the manufacturers' protocols as follows. 100 μl of BALF was added to equal volumes of the recommended reagents and incubated for either 30 min (LDH) or 5 min (Albumin) and absorbance measured at 490 nm (LDH assay) or 630 nm (Albumin assay).
For in situ hybridization experiments, gene-specific riboprobes were synthesized by in vitro transcription using a Maxiscript SP6/T7 kit (Ambion) and unincorporated nucleotides were removed using RNA Mini Quick Spin columns (Roche). Paraffin embedded lung sections were pretreated as described [74], following deparaffinization in xylenes and rinsing in ethanol. In situ hybridization with 35 S-UTP-labeled riboprobes (antisense or control sense) was performed as described previously [75] with 0.1M dithiothreitol included in the hybridization mix. Hybridizations were performed at 50°C overnight. After stringent washing to remove unbound riboprobes, tissue sections were coated with NTB emulsion (Kodak) and exposed at 10°C for 14 days. Parallel hybridizations using the control sense riboprobes did not give rise to specific autoradiographic signals.
For supplementation experiments, infected mice were administered 50 ng of recombinant rCXCL1 (Cat. No. 573702, BioLegend), rCXCL2 (Cat. No. 582502) or rCXCL5 (Cat. No. 573302) in 50 μl PBS or PBS alone via the i.t. route (at 4 h p.i.) and monitored for survival or assayed for neutrophil recruitment. 50 ng of recombinant rCXCL1 was analyzed using the Limulus amoebocyte lysate assay kit (Lonza, Cat. No. 50647U) and endotoxin content was below the level of detection (0.1 EU/ml).
In vitro Aspergillus-BMM co-culture assays BMM stimulation with A. fumigatus germlings was performed as described in [23]. Briefly, BMMs derived from WT and CARD9 (−/−) mice were stimulated with UV-inactivated germlings (MOI = 1) for 18 h. Chemokine levels in culture supernatants were measured using ELISA kits from R&D Systems.

Statistical analysis
All results are expressed as mean (±SEM) derived from 3 independent experiments, unless stated otherwise. A Mann-Whitney U test was used for unpaired two groups or Wilcoxon signedrank test for paired two groups comparison. Non-parametric Kruskal-Wallis one-way analysis of variance followed by Dunn's post test was used for three groups comparison. Survival data was analyzed by long rank test unless stated otherwise. All statistical analyses were performed with GraphPad Prism software, v6.0c. A p value < 0.05 was considered significant and indicated with an asterisk. Neta Shlezinger for reviewing the manuscript, and staff at Experimental Histopathology (FHCRC) for help with histology slides preparation.