Neutrophil phagocyte oxidase activity controls invasive fungal growth and inflammation in zebrafish

Chronic granulomatous disease (CGD) is an inherited immunodeficiency caused by mutations in any of the five subunits CYBA, NCF4, NCF1, NCF2 and NOX2 (hereafter referred to as p22^phox, p40^phox, p47^phox, p67^phox and p91^phox) comprising the phagocyte NADPH oxidase (PHOX) complex (Bedard and Krause, 2007). It is characterized by inflammatory disorders, granuloma formation, and recurrent or chronic bacterial and fungal infections, such as invasive aspergillosis (King et al., 2016; Levine et al., 2005). While Aspergillus is typically innocuous to immunocompetent individuals, susceptibility to invasive aspergillosis is increased in people with inherited immunodeficiency or medically-induced immunosuppression that impairs innate immune cell function, especially those with neutropenia, i.e. neutrophil deficiency (King et al., 2016; Segal et al., 2010). Relative to other inherited immune disorders, CGD is the most significant predisposing condition for developing invasive aspergillosis (Blumental et al., 2011), and invasive aspergillosis is responsible for many of the infection-related mortalities of CGD patients (Henriet et al., 2012; Marciano et al., 2015), with A. fumigatus as the primary causative agent (Marciano et al., 2015). Despite its rare occurrence in other immunocompromised populations, A. nidulans infection is the second most frequent Aspergillus infection in CGD, and is associated with higher morbidity and mortality rates than A. fumigatus (Henriet et al., 2012; Gresnigt et al., 2018). However, the underlying interactions between A. nidulans and the innate immune response that contribute to the unique susceptibility of CGD patients to A. nidulans infections remain unknown.

The pathological consequences of CGD are underpinned by the inability of CGD phagocytes to produce reactive oxygen species (ROS). Phagocytic ROS has been reported to have both direct microbicidal and immunomodulatory functions but which of these functions provide the dominant mechanism of host defense against Aspergillus remains unclear. In vitro, neutrophil-derived ROS can damage Aspergillus (Rex et al., 1990) but Aspergillus can counter oxidative stress through its own antioxidant pathways (Chang et al., 1998; Lambou et al., 2010; Wiemann et al., 2017). In vivo, mouse models of CGD are susceptible to Aspergillus infection, with both an impaired host defense and altered inflammatory response (Bonnett et al., 2006; Cornish et al., 2008; Morgenstern et al., 1997; Pollock et al., 1995). Furthermore, loss of PHOX activity causes aberrant inflammation in response to sterile injury, supporting a major role of PHOX in regulating inflammation independent of microbial control (Bignell et al., 2005; Morgenstern et al., 1997).

There is also a gap in understanding the cell-specific contribution of macrophage- and neutrophil-derived ROS in the innate immune response to Aspergillus infection. Host defense against inhaled Aspergillus conidia is mediated by the innate immune system (Balloy and Chignard, 2009), and in an immunocompetent host both macrophages and neutrophils can kill conidia or inhibit germination (Jhingran et al., 2012; Zarember et al., 2007), while neutrophils are primarily responsible for the destruction of hyphae post germination (Diamond et al., 1978; Gazendam et al., 2016; Knox et al., 2014). There are varying reports on the importance of macrophage ROS for inhibiting conidial germination but there is consistent support for the role of macrophage ROS in regulating cytokines involved in neutrophil recruitment (Cornish et al., 2008; Grimm et al., 2013). In neutrophils, PHOX activity is dispensable for inhibiting germination in vitro (Zarember et al., 2007) but plays a minor role in germination inhibition and killing of Aspergillus spores in vivo (Cornish et al., 2008; Jhingran et al., 2012). Neutrophil ROS is also involved in regulating pro-inflammatory signals, and neutrophils from CGD patients have increased expression of pro-inflammatory cytokines at a basal level as well as in response to pathogen challenge (Kobayashi et al., 2004; Smeekens et al., 2012). Together, the variable requirement of PHOX activity in restricting Aspergillus growth and the increased production of pro-inflammatory cytokines by CGD phagocytes support a role for PHOX activity in modulating neutrophil inflammation commonly associated with CGD. However, how neutrophil-specific ROS mediates fungal growth, fungal clearance, inflammation and host survival in response to A. nidulans infection remains unclear.

In this study, we aimed to determine the neutrophil-specific role of PHOX in both clearing fungal burden and controlling inflammation, and to identify characteristics of A. nidulans that allow it to cause disease specifically in CGD hosts by using an established Aspergillus-larval zebrafish infection model (Herbst et al., 2015; Knox et al., 2014; Rosowski et al., 2018). The transparency of zebrafish larvae, the availability of mutant and transgenic lines with fluorescently labeled phagocytes, and cell-specific rescues allows us to observe both fungal burden and host inflammation during infection over the course of several days. Furthermore, zebrafish express functional PHOX in both neutrophils and macrophages (Brothers et al., 2011; Niethammer et al., 2009; Yang et al., 2012), making the larval zebrafish an attractive model for studying the cell-specific roles of ROS in progression of infection.

We demonstrate that PHOX-deficient larvae (p22^{phox−/− (sa11798)}) have increased susceptibility to A. nidulans infection, similar to the susceptibility of human CGD patients. Live imaging reveals that, in a wild-type host, A. nidulans germinates faster, evokes a stronger immune response and is cleared sooner after infection, compared to A. fumigatus. Global PHOX activity does not prevent conidial germination but reduces extensive invasive growth of A. nidulans hyphae and prevents excessive neutrophil recruitment to the infection site. Restoring PHOX activity in just neutrophils limits the area of invasive fungal growth, and fully restores neutrophil recruitment and larval survival to those of wild-type levels. Our data demonstrate that A. nidulans elicits a distinct immune response that leads to both greater inflammation and hyphal-induced tissue damage in PHOX-deficient hosts, and that neutrophil-derived ROS can limit both invasive fungal growth and inflammation.

To develop a zebrafish model of CGD, we began by characterizing embryos containing the p22^phox allele sa11798 from the Sanger Zebrafish Mutation Project. This allele has a nonsense mutation that is predicted to cause the loss of one of three transmembrane domains of the p22 protein (Fig. 1A) and will be referred to as p22^−/− herein. p22^−/− larvae have decreased p22 protein levels, as determined by western blot analysis (Fig. 1B, Fig. S1). p22^−/− larvae also show reduced survival in response to both A. fumigatus and A. nidulans, supporting its use as a model of CGD. Interestingly, A. nidulans caused more death than A. fumigatus in p22^−/− larvae, with a hazard ratio of 74.2 over infection in control larvae, compared to just 7.3 for A. fumigatus, demonstrating a specifically increased susceptibility to A. nidulans infection (Fig. 1C). This specific susceptibility to A. nidulans was also observed in p22^phox knockdown larvae (p22 morpholino) (Fig. S2). To determine whether this A. nidulans-induced death is unique to a PHOX-deficient host, we also compared the ability of A. nidulans and A. fumigatus to cause disease in a generally immunosuppressed host. Wild-type larvae were infected with A. nidulans or A. fumigatus and treated with the glucocorticoid and general immunosuppressant dexamethasone or a DMSO vehicle control immediately following infection. In contrast to the p22 mutant, A. nidulans caused fewer deaths among these dexamethasone-treated larvae than A. fumigatus (Fig. 1D). Together, these data demonstrate the specific increase in susceptibility to A. nidulans in p22 mutant larvae, providing a powerful tool to study the role of phagocyte ROS in specific host–pathogen interactions.

Using live imaging, we first wanted to determine how fungal growth and host responses differ between A. nidulans and A. fumigatus infections in wild-type hosts. Germination of Aspergillus conidia to tissue-invasive hyphae is a crucial driver of Aspergillus pathogenesis that triggers immune activation and fungal clearance by the host (Hohl et al., 2005; Rosowski et al., 2018). To determine whether A. nidulans and A. fumigatus have different growth kinetics in vivo, we measured spore germination and hyphal growth in larval zebrafish. A. fumigatus or A. nidulans spores expressing red fluorescence protein (RFP) were injected into the hindbrain of wild-type larvae, and imaged by confocal microscopy at 6, 24, 48 and 72 hours post infection (hpi) to track Aspergillus germination and growth over the course of infection within individual larvae (Fig. 2A,B). All A. nidulans-infected larvae contained germinated spores at 24 hpi but the percentage of larvae with persistent germinated spores decreased to 84% at 48 hpi and 47% at 72 hpi, suggesting clearance of A. nidulans by the host immune system (Fig. 2C). In contrast, A. fumigatus germination was delayed until 48 hpi but the percentage of larvae with germination continued to increase up to 72 hpi with limited evidence of fungal clearance (Fig. 2C). In addition to germination, we monitored invasive fungal growth as defined by the presence of lateral hyphal branching. The pattern of invasive growth followed a trend similar to that of germination. A. nidulans-infected larvae experienced faster development of invasive fungal growth and subsequent clearance relative to larvae infected with A. fumigatus (Fig. 2D). While there was variability in fungal growth across individual larvae, the trend of early growth of A. nidulans followed by clearance was consistent, while all growth of A. fumigatus occurred later (Fig. 2B).

We further investigated differences in immune clearance of A. fumigatus and A. nidulans by quantifying the fungal burden in individual larvae through colony-forming unit (CFU) plating experiments. Wild-type larvae were infected with A. fumigatus or A. nidulans spores and we measured CFUs at day post infection (dpi) 0, 1, 3 and 5. A large percentage of the A. fumigatus spore burden persisted in the host up to 5 dpi, consistent with previous findings (Knox et al., 2014) while the A. nidulans spore burden dropped dramatically by 1 dpi (Fig. 2E), providing further evidence that A. nidulans is cleared sooner after infection than A. fumigatus in wild-type larvae.

Germination of Aspergillus reveals cell wall polysaccharides on the hyphal surface that contribute to immune activation (Henriet et al., 2016; Hohl et al., 2005). We hypothesized that differences in germination rate between A. nidulans and A. fumigatus would, therefore, result in differences in the inflammatory response following infection. To analyze NF-κB activation induced by A. nidulans and A. fumigatus infections in vivo, we utilized an NF-κB activation reporter line as a visual proxy of pro-inflammatory activation (NF-κB RE:EGFP; Kanther et al., 2011). At 1 dpi, A. nidulans-infected larvae had significantly more NF-κB reporter activity and a significantly higher fungal burden due to hyphal growth compared to A. fumigatus-infected larvae (Fig. 3A-C). These data indicate that A. nidulans elicits a stronger host immune response early in infection as compared to A. fumigatus, which is likely to contribute to its rapid clearance in a wild-type host.

We next determined whether the growth kinetics of A. nidulans are altered in p22^−/− larvae in order to address why a PHOX-deficient host is more susceptible to A. nidulans. We infected p22^−/− and p22^+/− control larvae with RFP-expressing A. nidulans and imaged the larvae by confocal microscopy at 6, 24, 48, 72 and 96 hpi (Fig. 4A-E, Fig. S3, Movies 1-2). In both p22^−/− and p22^+/− backgrounds, A. nidulans spore germination occurred by 24 hpi in 100% of larvae, with similar kinetics (Fig. 4A-C). However, invasive hyphal growth was significantly increased, and fungal clearance was impaired by 48 hpi in p22^−/− larvae (Fig. 4B-E). These data suggest a role for p22^phox in limiting hyphal growth post germination rather than inhibiting the initiation of germination.

In addition to fungal burden, PHOX can also modulate inflammatory cell recruitment. Because the primary innate immune cell recruited to hyphae are neutrophils (Diamond et al., 1978; Gazendam et al., 2016; Knox et al., 2014), we also imaged neutrophil infiltration over the course of A. nidulans infection in a labeled-neutrophil line (mpx:gfp) crossed with the p22^phox mutant. A. nidulans recruited significantly more neutrophils in p22^−/− larvae as compared to p22^+/− larvae at 48, 72 and 96 hpi (Fig. 4F, Fig. S3, Movies 1-2). Whereas neutrophil recruitment in control hosts peaks at 24 hpi, recruitment in p22^−/− larvae continues to increase at least up to 96 hpi and correlates with the presence of invasive hyphal growth (Fig 4D).

Previous studies have reported that PHOX-deficient hosts have increased expression of inflammatory cytokines (Henriet et al., 2012; Smeekens et al., 2012). To determine whether there is increased cytokine and chemokine expression in p22^−/− larvae during steady-state, i.e. uninfected, conditions, we performed RT-qPCR analyses of the zebrafish cytokines il1b, il6 and tnfa, and chemokines cxcl8-l1 and cxcl8-l2 (the homologs of mammalian Cxcl8) in uninfected p22^+/+ and p22^−/− larvae at 3 dpf. The p22^−/− larvae exhibited variable levels of cytokine expression, with a trend towards increased expression of il6 and tnfa in pooled larvae compared to the wild-type control (Fig. 5A). We next determined whether there is also increased expression of inflammatory genes during A. nidulans infection in p22^+/+ and p22^−/− larvae at 24 hpi. The p22^−/− larvae exhibited statistically significant upregulation of tnfa, and a trend towards increased expression of il1b, il6 and cxcl8-l2 in individual larvae compared to controls (Fig. 5B,C). Taken together, our findings indicate that there are basic differences between uninfected p22^−/− and p22^+/+ larvae. Moreover, levels of inflammatory mediators were increased in p22^−/− larvae compared with those in p22^+/+ larvae following infection with A. nidulans.

Our findings demonstrate increased neutrophil inflammation in response to A. nidulans in PHOX-deficient larvae, suggesting that neutrophils may play a key role in the pathogenesis of this infection. To further address the role of neutrophils and neutrophil ROS, we utilized two different larval zebrafish models of human neutrophil deficiency. One transgenic zebrafish line – for WHIM syndrome – models congenital neutropenia by expressing GFP-tagged truncated CXCR4 (mpx:CXCR4-WHIM-GFP), the other – modeling a leukocyte adhesion deficiency – expresses a dominant-negative Rac2D57N mutation in neutrophils (mpx:rac2D57N). Although both models have impaired neutrophil migration to sites of infection (Deng et al., 2011; Walters et al., 2010), only the Rac2D57N model has also neutrophils that have an impaired oxidative burst, i.e. release of ROS (Ambruso et al., 2000; Williams et al., 2000). Larvae with CXCR4 or RacD57N mutations (hereafter referred to as WHIM or Rac2D57N larvae, respectively) showed impaired host survival in response to either A. nidulans or A. fumigatus. Although there was no difference in the survival of WHIM larvae infected with A. fumigatus compared to those infected with A. nidulans (Fig. 6A), we did observe decreased survival of Rac2D57N larvae infected with A. nidulans compared to those infected with A. fumigatus (Fig. 6B). These data suggest that neutrophil-derived ROS specifically mediates survival in response to A. nidulans infection.

We next tested tested whether PHOX-derived ROS production in neutrophils alone is sufficient to mediate survival of p22^−/− larvae. For this, p22^−/−; mpx:gfp control larvae or p22^−/− neutrophil rescue larvae that re-express functional p22^phox exclusively in neutrophils (p22^−/−; mpx:p22:gfp) were infected with RFP-expressing A. nidulans and imaged by confocal microscopy at 6, 24, 48, 72 and 96 hpi (Figs 7A-F and 8A). The p22^−/− larvae with p22^phox re-expressed just in neutrophils had a similar burden of germinated A. nidulans spores but showed less-extensive fungal growth than p22^−/− larvae, suggesting that neutrophil production of ROS limits the area of invasive fungal growth (Figs 7A-E and 8A). Furthermore, neutrophil-specific expression of functional p22^phox almost completely restored wild-type-like neutrophil recruitment to infection, suggesting that neutrophil-produced ROS can also limit excessive neutrophil inflammation (Fig. 7F, Fig. S4). Because rescue of p22^phox in neutrophils alone improved both fungal burden and neutrophil inflammation in p22^−/− larvae, we next tested whether neutrophil-specific p22^phox expression improved the overall host survival. Indeed, we found that neutrophil-specific expression of functional p22^phox rescued host survival back to wild-type levels (Fig. 8B). These data demonstrate that neutrophil p22^phox contributes to clearance of hyphae and is sufficient for regulation of excessive neutrophil recruitment. Control of both fungal growth and excessive inflammation allows restoration of larval survival and suggests an important role of neutrophil-specific p22^phox in controlling A. nidulans infections.

Here, we developed a PHOX-deficient zebrafish model to image the temporal and spatial dynamics of A. nidulans infection in a live, intact host. Few studies have addressed the unique susceptibility to the typically avirulent fungus A. nidulans in CGD. We found that A. nidulans is significantly more virulent than A. fumigatus in p22^−/− larvae – although not in a generally immunocompromised or neutropenic host. Interestingly, A. fumigatus caused only a low level of death in p22^−/− larvae despite being the most frequently isolated fungus from CGD patients (King et al., 2016). The relatively weak virulence of A. fumigatus could be a result of strain variation, which can have a significant impact on host survival (Amarsaikhan et al., 2014; Knox et al., 2016; Kowalski et al., 2016; Rizzetto et al., 2013; Rosowski et al., 2018). Remarkably, rescue of p22^phox function specifically in neutrophils was sufficient to reduce invasive fungal growth but did not affect spore germination. It also fully restored neutrophil recruitment and host survival to wild-type levels, suggesting that death of p22^−/− larvae is a combined result of unconstrained fungal growth and of damaging inflammation that is normally regulated by neutrophil production of ROS. To our knowledge, this is the first report of neutrophil ROS regulating host susceptibility in response to a specific fungal pathogen.

By using non-invasive live imaging techniques, we tracked germination and hyphal growth over the course of infection, and observed that A. nidulans germinates faster and develops invasive hyphae faster than A. fumigatus, both of which are likely to contribute to the clearance of A. nidulans in a wild-type host. Germination reveals immunogenic carbohydrates on the hyphal surface, which induce phagocyte response to infection (Hohl et al., 2005). Moreover, we have recently reported that faster germination and hyphal growth can increase fungal clearance by increasing immune activation and neutrophil-mediated killing (Rosowski et al., 2018). Similarly, we observed stronger NF-κB activation from A. nidulans than from A. fumigatus at 24 hpi, concomitantly with higher levels of germination and fungal growth, which is consistent with other reports in that A. nidulans stimulates the immune system more than A. fumigatus (Gresnigt et al., 2018; Henriet et al., 2016; Smeekens et al., 2012). Considering this increased immune activation and the fact that CGD is defined by chronic inflammation, we hypothesize that, in a wild-type host, this heightened immune response translates into rapid fungal clearance; but in a PHOX-deficient host it results in host-damaging inflammation.

Here, we observed both extensive hyphal growth and massive neutrophil recruitment in p22^phox-deficient larvae infected with A. nidulans. This excessive neutrophil recruitment of p22^phox-deficient larvae in response to A. nidulans infection is consistent with previous reports that PHOX activity regulates neutrophil inflammation during infection in mice and zebrafish (Mesureur et al., 2017; Segal et al., 2010). Excessive neutrophil infiltration could contribute to host death, even in the absence of a functional PHOX complex, as neutrophils can damage host tissues by amplifying the response of other immune cells or by releasing enzymes and antimicrobial proteins that directly harm host tissue (Wang, 2018).

How does ROS regulate neutrophil recruitment and inflammation? We observed increased expression of pro-inflammatory cytokines in p22^phox-deficient larvae under steady-state conditions and in response to A. nidulans, recapitulating the enhanced cytokine production seen in human CGD patients (Kobayashi et al., 2004; Smeekens et al., 2012; Warris et al., 2003). We found consistent upregulation of tnfa, and a trend towards increased expression of il6 and cxcl8-l2 (Cxcl8 in mammals) at 24 hpi, similar to observations in CGD mice (Morgenstern et al., 1997). TNF and IL6 mediate response to infection by promoting neutrophil recruitment (Hind et al., 2018), and ROS production (Figari et al., 1987). Additionally, CXCL8 is a strong neutrophil chemoattractant in humans (Hoffmann et al., 2002) and zebrafish (de Oliveira et al., 2013). Although the observed increase in pro-inflammatory cytokine expression cannot be directly linked to neutrophil recruitment, it is interesting to note that there are two homologs of human CXCL8 in zebrafish – Cxcl8-l1 and Cxcl8-l2. Both homologs can have distinct or redundant roles depending on the stimulus, including mediation of neutrophil recruitment versus resolution of inflammation (de Oliveira et al., 2013, 2015). Further studies regarding the effects of these two isoforms by using our zebrafish CGD model might provide insight into whether the hyperinflammation in this CGD model is due to enhanced neutrophil recruitment or impaired resolution.

Additionally, the specific susceptibility of Rac2D57N – but not WHIM – larvae to A. nidulans suggests that neutrophil ROS act as a long-range signal to mediate inflammation, because neutrophils in WHIM larvae are unable to migrate to sites of infection (Walters et al., 2010) but can still produce ROS. Considering that ROS are able to diffuse through tissues (Niethammer et al., 2009), and the close association of macrophages and neutrophils during Aspergillus infection (Knox et al., 2017; Rosowski et al., 2018), it is possible that neutrophil-derived ROS contribute to resolution of inflammation during infection by influencing macrophage behavior. We have previously reported that zebrafish larvae deficient in p22^phox show hyperinflammation and defective neutrophil resolution at the site of sterile injury. This is partially due to impaired neutrophil reverse migration mediated by macrophages (Tauzin et al., 2014), demonstrating a signaling role for PHOX in regulating inflammation. We must also consider the role of ROS production by other cell types, such as macrophages. It is known that macrophage-derived ROS contributes to neutrophil recruitment during Aspergillus infection, and that restoration of PHOX function in macrophages provides wild-type resistance in a murine model of invasive aspergillosis (Cornish et al., 2008; Grimm et al., 2013). Therefore, further investigation is needed to establish whether neutrophil- or macrophage-specific ROS have independent or redundant roles in protection against Aspergillus infection.

ROS have also been implicated as a direct microbial killer, but the ability of active PHOX to eradicate Aspergillus has been debated. One argument against the importance of ROS in direct microbial killing in this specific interaction between CGD hosts and A. nidulans is the finding that the latter, as compared to A. fumigatus, is not susceptible to killing through neutrophil-derived ROS in vitro (Henriet et al., 2011). In our experiments, we did observe a significantly greater hyphal burden of A. nidulans in p22^−/− compared to p22^+/− larvae. However, the increase in inflammation and resulting tissue damage in these larvae might also yield an environment preferable for growth of A. nidulans, making it difficult to dissect these two roles of PHOX activity.

Our data support a larval zebrafish model with conserved p22^phox activity that is crucial for defense against infection with A. nidulans, and contributes to host survival by mediating both fungal growth and damaging inflammation. Reconstitution of p22^phox in neutrophils supported the idea of cell-specific roles for PHOX activity during infection; however, the full extent of phagocyte contributions to inflammation and fungal clearance warrants further investigation. For now, we think the full rescue of neutrophil recruitment and the partial rescue of invasive growth during A. nidulans infection contributes to host survival in two ways: (1), by decreasing non-oxidative host tissue damage through neutrophils; (2) by decreasing host tissue damage through invasive hyphae. Future work is needed to clarify whether the neutrophilic inflammation observed in zebrafish larvae deficient for p22^phox damages host tissue and, if so, through what mechanisms. By exploiting the optical transparency and transgenic resources of the larval zebrafish CGD model, we and others are well-positioned to further investigate the mechanisms that mediate PHOX-dependent regulation of inflammation as well as host survival.

Animal care and use protocol M005405-A02 was approved by the University of Wisconsin-Madison College of Agricultural and Life Sciences (CALS) Animal Care and Use Committee. This protocol adheres to the federal Health Research Extension Act and the Public Health Service Policy on the Humane Care and Use of Laboratory Animals, overseen by the National Institutes of Health (NIH) Office of Laboratory Animal Welfare (OLAW).

Adult zebrafish and larvae were maintained as described previously (Knox et al., 2014). Larvae were anesthetized prior to experimental procedures in E3 water containing 0.2 mg/ml Tricaine (ethyl 3-aminobenzoate, Sigma). To prevent pigment formation in larvae during imaging experiments, larvae were maintained in E3 containing 0.2 mM N-phenylthiourea beginning 1 day post fertilization (dpf) (PTU, Sigma Aldrich). All zebrafish lines used in this study are listed in Table S1.

Zebrafish containing the p22^phox allele sa11798 were isolated through the Sanger Zebrafish Mutation Project, Wellcome Sanger Institute, and obtained from the Zebrafish International Resource Center (ZIRC). The sa11798 point mutation was detected using primers designed for using the dCAPS method (Neff et al., 1998). Forward primer: p22_mutFor_MseI_F: 5′-CTTTTGGACCCCTGACCAGAAATTA-3′. Reverse primer: p22_mutFor_MseI_R: 5′-TGGCTAACATGAACCCTCCA-3′. The 211 bp PCR product was digested with MseI and analyzed on 3% agarose gel, with detected band sizes of: +/+ (115 bp, 96 bp), +/− (115 bp, 96 bp, 74 bp) and −/− (115 bp, 74 bp). The neutrophil p22^phox rescue line was genotyped with the same protocol as above, but the restriction pattern from these larvae contains an additional 84 bp product in all genotypes. For easier distinction of bands, the genotypes of these larvae were determined by analyzing digest products on 3% MetaPhor agarose gel. Neutrophil-specific restoration of p22^phox was achieved by using the Tg(mpx:cyba:gfp) line which expresses p22^phox and green fluorescence protein (GFP) under the neutrophil-specific mpx promoter (Tauzin et al., 2014). The Tg(mpx:cyba:gfp) line was then crossed to the p22^phox sa11798 allele line and larvae with GFP+ neutrophils were selected to grow up to establish a stable line. The Tg(mpx:cyba:gfp); p22^phox−/− is also referred to as p22^−/−; mpx:p22:gfp in the results, for simplicity.

For p22^phox knockdown (Fig. S2), 3 nl of cyba splice-blocking MO solution (Gene Tools; ZFIN MO1-cyba: 5′-ATCATAGCATGTAAGGATACATCCC-3′) was injected into embryos at the one- to four-cell stage at a concentration of 500 µM (Tauzin et al., 2014).

All Aspergillus strains used in this study are listed in Table S2. All strains were grown on solid glucose minimal medium (GMM). A. fumigatus was grown at 37°C in darkness and A. nidulans was grown at 37°C in constant light to promote asexual conidiation. Conidial suspensions for microinjection were prepared using a modified protocol from Knox et al. (2014) that includes an additional filtration step to eliminate hyphal fragments and conidiophores of A. nidulans from the suspension. For consistency, the additional filtration step was also used for isolation of A. fumigatus conidia. Aspergillus was grown for 3–4 days on solid GMM after having been plated at a concentration of 1×10⁶ conidia/10 cm plate. Fresh conidia were harvested in 0.01% Tween water by scraping with an L-spreader. The spore suspension was then passed through sterile Miracloth into a 50 ml conical tube and the volume was adjusted to 50 ml with 0.01% Tween. The spore suspension was centrifuged at 900 g for 10 min at room temperature and the spore pellet was re-suspended in 50 ml 1× PBS. The spore suspension was then vacuum filtrated using a Buchner filter funnel with a glass disc containing 10–15 µm diameter pores. The filtered suspension was centrifuged at 900 g for 10 min and re-suspended in 1 ml 1× PBS. Conidia were counted using a hemacytometer and the concentration was adjusted to 1.5×10⁸ spores/ml. Conidial stocks were stored at 4°C and used up to 1 month after harvesting.

For western blotting, 50–100 2-dpf larvae were pooled and de-yolked in Ca²⁺-free Ringer's solution with gentle disruption with a p200 pipette. Larvae were washed twice with PBS and stored at −80°C until samples were lysed by sonication in 20 mM Tris pH 7.6, 0.1% Triton X-100, 0.2 mM phenylmethylsulfonyl fluoride (PMSF), 1 μg/ml Pepstatin, 2 μg/ml Aprotinin and 1 μg/ml Leupeptin at 300 μl per 100 larvae while on ice and clarified by centrifugation. Protein concentrations were determined using a bicinchoninic acid protein assay kit (Thermo Fisher Scientific), according to the manufacturer's instructions. Total protein (∼50 µg) was loaded on 6–20% gradient SDS-polyacrylamide gels and transferred to nitrocellulose. Zebrafish p22^phox was detected using an antibody against full-length human p22^phox [p22-phox (FL-195): sc-20781, Santa Cruz Biotechnology, 1:500 dilution]. Western blots were imaged with an Odyssey Infrared Imaging System (LI-COR Biosciences).

Anesthetized larvae (at 2 dpf) were microinjected with conidia into the hindbrain ventricle via the otic vesicle as described by Knox et al. (2014). 1% Phenol Red was mixed at 1:2 ratio with the conidial suspension, so the inoculum was visible in the hindbrain after injection. After infection larvae were rinsed 3× with E3 without Methylene Blue (E3-MB) to remove the Tricaine solution and were then transferred to individual wells of a 96-well plate for survival experiments. For imaging experiments larvae were kept in 35-mm dishes or 48-well plates. Survival was checked daily for 7 days and larvae with a heartbeat were considered to be alive. We aimed for an average spore dose of 60 spores and the actual spore dose for each experiment was monitored by CFU counts and reported in the figure legends.

To quantify the initial spore dose and fungal burden in survival and fungal clearance assays, anesthetized larvae were collected immediately after spore injection and placed in individual 1.5 ml microcentrifuge tubes in 90 µl of 1×PBS with 500 µg/ml kanamycin and 500 µg/ml gentamycin. Larvae were homogenized using a mini-bead beater for 15–20 s and the entire volume of the tube was plated on GMM solid medium. Plates were incubated at 37°C for 2–3 days and colony-forming units (CFUs) were counted. At least 7–8 larvae were used for each condition, time point and replicate. To quantify the percentage of initial spore burden, the number of CFUs was normalized to the average CFU count on day 0.

We used dexamethasone (Sigma) to induce general immunosuppression of wild-type larvae. Dexamethasone was re-constituted to 10 mM with DMSO for a stock solution and stored at −20°C. Immediately following spore microinjection, larvae were treated with 10 µM dexamethasone or DMSO (0.01%); larvae remained in the drug bath for the entirety of the survival experiment.

For analysis of pooled larvae (Fig. 5A), RNA was extracted from pools of ten whole larvae with 500 µl TRIzol reagent (Invitrogen). For analysis of individual larvae (Fig. 5B,C), RNA was extracted from individual whole larvae with 100 µl TRIzol reagent (Invitrogen). cDNA was synthesized with SuperScript III RT and oligo-dT (Invitrogen). cDNA was used as the template for quantitative PCR (qPCR) using FastStart Essential Green DNA Master (Roche) and a LightCycler96 (Roche). Data were normalized to rps11 within each sample using the ΔΔCq method (Livak and Schmittgen, 2001). Data in 5A-B were then normalized to those of the wild-type condition for each cytokine. The mean of ΔΔCq values across multiple individual larvae was used to calculate the fold-changes displayed in Fig. 5B. Fold-changes in B, therefore, represent pooled data collected from 33 individual p22^−/− and 30 p22^+/+ individual larvae over three experiments. All qPCR primers are listed in Table S3.

Pre-screening of larvae with fluorescent markers was performed on a zoomscope (EMS3/SyCoP3; Zeiss; Plan-NeoFluar Z objective). Multi-day imaging experiments were performed on a spinning disk confocal microscope (CSU-X; Yokogawa) with a confocal scanhead on a Zeiss Observer Z.1 inverted microscope, Plan-Apochromat NA 0.8/20x objective, and a Photometrics Evolve EMCCD camera. Images were acquired using ZEN software (Zeiss). Larvae imaged at multiple time points were kept in 48-well plates in E3-MB with PTU. For imaging, larvae were removed from the 48-well plate and anesthetized in E3-MB with Tricaine before being loaded into zWEDGI chambers (Huemer et al., 2017). Larvae were positioned so that the hindbrain was entirely visible during imaging. All z-series images were acquired in 5 µm slices. After imaging, the larvae were rinsed with E3-MB to remove Tricaine and were returned to the 48-well plate with E3-MB with PTU. To image the EGFP signal in the Tg(NF-κB RE:EGFP) line, larvae were anesthetized and positioned at the bottom of a glass-bottom dish in 1% low-melting point agarose. Images were acquired on a laser-scanning confocal microscope (FluoView FV1000; Olympus) with an NA 0.75/20x objective and FV10-ASW software (Olympus).

All displayed images in Figs 1–7 represent maximum intensity projections of z-series images that were generated using Imaris software. The brightness and contrast were adjusted for displayed images also using Imaris. For analysis of germination and invasive growth, all images were viewed as z-stacks and maximum intensity projections in Zen software (Zeiss) to score the presence of germination and hyphae. To measure the EGFP signal in Tg(NF-κB RE:EGFP) larvae, images were analyzed in Fiji. A single z-slice from the middle of the hindbrain was used to measure EGFP signal intensity. The measurement area and integrated density of the EGFP signal measured inside a region of interest (ROI) encompassing the hindbrain, with the ROI determined manually from the brightfield image. No alterations were made to the images prior to analysis. 2D fungal area was measured by manually thresholding the red fluorescence protein (RFP) signal from A. nidulans and measuring the area of RFP signal. Fungal area measurements were made from maximum intensity projections of z-stacks in Fiji. Neutrophil recruitment during infection was analyzed by manually counting neutrophils within the hindbrain and neutrophils in contact with hyphae in cases where hyphae grew outside of the hindbrain. In addition to counting cells, neutrophil recruitment was also analyzed by manually thresholding the GFP signal from neutrophils and measuring the area of GFP signal within the hindbrain and neutrophils in contact with hyphae in cases where hyphae grew outside of the hindbrain. Both counting and GFP-signal area measurements were made from maximum intensity projections of z-stacks in Fiji. Movies 1 and 2 were prepared, and brightness and contrast of each channel were adjusted in Fiji. Depth-encoded maximum intensity projections of z-stacks were generated in Fiji.

All experiments and statistical analyses represent at least three independent replicates. The number of replicates for each experiment is indicated in the figure legends. Survival experiments were analyzed using Cox proportional hazard regression analysis with the experimental condition included as a group variable, as previously described (Knox et al., 2014). The pair-wise P values, hazard ratios and experimental n values are displayed in each figure or figure legend. For analysis of spore germination and invasive growth pair-wise comparisons were made using Student's t-tests; data represent means±standard deviation (±s.d.). Integrated densities of GFP/EGFP signal, RFP signal, neutrophil counts and spore burden represent least-squared adjusted means±standard error of the mean (LSmeans±s.e.m.) and were compared by ANOVA with Tukey's multiple comparisons. Fold-changes in gene expression in p22^−/− larvae were compared to the normalized control value of 1 using one-sample t-tests. Cox proportional hazard regression, least-squared adjusted means and ANOVA analyses were performed using R version 3.4.4. t-tests and graphical representations were done in GraphPad Prism version 7.

We thank members of the Huttenlocher and Keller labs for helpful discussions of the research and manuscript. We thank Jens Eickhoff (Department of Biostatistics and Medical Informatics, University of Wisconsin-Madison) for guidance on and assistance with statistical analyses.