A Host-Pathogen Interaction Screen Identifies ada2 as a Mediator of Candida glabrata Defenses Against Reactive Oxygen Species

Candida glabrata (C. glabrata) forms part of the normal human gut microbiota but can cause life-threatening invasive infections in immune-compromised individuals. C. glabrata displays high resistance to common azole antifungals, which necessitates new treatments. In this investigation, we identified five C. glabrata deletion mutants (∆ada2, ∆bas1, ∆hir3, ∆ino2 and ∆met31) from a library of 196 transcription factor mutants that were unable to grow and activate an immune response in Drosophila larvae. This highlighted the importance of these transcription factors in C. glabrata infectivity. Further ex vivo investigation into these mutants revealed the requirement of C. glabrata ADA2 for oxidative stress tolerance. We confirmed this observation in vivo whereby growth of the C. glabrata Δada2 strain was permitted only in flies with suppressed production of reactive oxygen species (ROS). Conversely, overexpression of ADA2 promoted C. glabrata replication in infected wild type larvae resulting in larval killing. We propose that ADA2 orchestrates the response of C. glabrata against ROS-mediated immune defenses during infection. With the need to find alternative antifungal treatment for C. glabrata infections, genes required for survival in the host environment, such as ADA2, provide promising potential targets.

model of infection where C. glabrata cells were found to cross the chlorio-allantoic membrane (CAM) (Jacobsen et al., 2011). In this context, a way in which C. glabrata may gain entry to host tissues is through pseudo-hyphal growth, which has been reported in vitro (Csank and Haynes 2000). In vivo, C. glabrata persists through endocytosis with hardly any host cell damage (Li et al., 2007). Because of this low host cell damage, the cytokine profile of C. glabrata-infected epithelia differs significantly from that of C. albicans-infected cells (Li et al., 2007). As a result, this leads to a strong neutrophil infiltration typical of C. albicans infection whereas infection with C. glabrata is associated with mononuclear cells (Westwater et al., 2007;Jacobsen et al., 2010). Once within the host, survival and the establishment of infection depends on the ability of C. glabrata to mount efficient responses to a changing, stressful environment; to acquire often limited nutrients; to evade the immune response.
The intrinsic resistance of C. glabrata to oxidative stress is of particular note (Cuéllar-Cruz et al., 2008;Kaloriti et al., 2012). The oxidative burst elicited by immune cells is a first line of defense against invading microorganisms (reviewed in Fang 2004). In turn, the ability of the pathogen to detoxify such reactive oxygen species (ROS), via the expression of detoxifying enzymes (catalase and superoxide dismutase) and production of antioxidants glutathione and thioredoxin, is important for surviving immune attack (reviewed in Brown et al., 2009). For example deletion of SKN7, a transcription factor mediating the oxidative stress response, attenuates C. glabrata virulence in a murine model of disseminated candidiasis (Saijo et al., 2010). This suggests additional mechanisms that contribute to increased oxidative stress resistance in C. glabrata beyond what we know from S. cerevisiae. Therefore, knowing the gene networks orchestrating C. glabrata responses to the host environment would be highly beneficial in understanding C. glabrata infection biology. Once C. glabrata has expanded to internal organs, its high resistance to common azole antifungals makes it hard to treat (Pfaller et al., 2010). High cost antifungals including echinocandins (e.g., caspofungin), anidulafungin and micafungin can also induce resistance (see Perlin 2007;Perlin 2014;Singh-Babak et al. 2012).
In light of the increased incidence of drug resistant C. glabrata infections, successful alternative treatment of C. glabrata perquisites further investigation of its pathogenicity. To this direction current mammalian models do not provide the most ideal approach to identify virulence factors because of their high cost, labor-intensive procedure, low statistical resolution and ethical concerns. Nonetheless, virulence factors in C. glabrata have been identified in these systems, including adhesins and cell-bound proteases (Cormack et al., 1999;Kaur et al., 2007;Tsoni et al., 2009;Seider et al., 2011). However, as large genomic libraries with deletion mutants become available in clinically relevant Candida pathogens, invertebrates provide an alternative approach for individually investigating virulence factors and host immune responses.
Drosophila is a small, inexpensive to rear and easy to manipulate model organism with a variety of sophisticated genetic tools and a fully annotated genome. Additionally, its interaction with Candida has been studied extensively, making it an excellent system for screening antifungals, investigating pathogenicity and identifying novel virulent factors in both systemic (injection) as well as gastrointestinal (feeding) infection (Brunke et al., 2015;Glittenberg et al., 2011a;Glittenberg et al., 2011b;Alarco et al., 2004;Chamilos et al., 2006).
The Drosophila Toll pathway is activated upon encountering a fungal infection (Lemaitre et al., 1996). Glucan-binding protein (GNBP3) binds to fungal b-1,3-glucan that activates a proteolytic cascade that results in cleavage of Spätzle (Gottar et al., 2006). Additionally, the protease Persephone (PSH) is also able to detect the activity of microbial virulence factors and can trigger the Toll pathway (Gottar et al., 2006). The signal is transmitted in the nucleus through the transcription factor, Dorsal-related immunity factor (DIF), which regulates targets genes including induction of the antimicrobial peptide (AMP) gene drosomycin (drs) (Rutschmann et al., 2000). An additional pathway in epithelial immunity is the Imd pathway (for immune deficiency). There, a transmembrane or intracellular peptidoglycan recognition protein (PGRP-LC and PGRP-LE respectively) forms a receptor-adaptor complex with IMD itself (a RIP1 homolog), which associates with FADD (the Fas-associated death domain protein) then recruiting the caspase-8 homolog DREDD (reviewed in Kounatidis and Ligoxygakis 2012). This receptor proximal complex acts as a platform for the recruitment and activation of the TAK1 kinase, which is presumed to phosphorylate the IkB-Kinase (ΙΚΚ). In turn, IKK (IKKa and IKKgotherwise known as Kenny) phosphorylates the N-terminal domain of the NF-kB homolog Relish, while DREDD cleaves the C-terminal (Paquette et al., 2010). N-terminal Rel is then free to move to the nucleus and regulate transcriptional targets including induction of antimicrobial peptide (AMP) genes (Stöven et al. 2000). In addition, Reactive Oxygen Species (ROS) are an important defense in controlling microbial invasion in the gut, regulated by the Drosophila dual oxidase gene (dDuox) (Ha et al., 2005).
Previous work from our laboratory involved Drosophila as a model organism to investigate pathogenicity of C. albicans following gastrointestinal (GI) infection (Glittenberg et al., 2011a;2011b). We determined that GI infection of wild type larvae with wild type C. albicans resulted in 83% of larvae reaching adulthood whereas 70% of larvae deficient for NF-kB-driven pathways developed into adults. In particular, the gut commensal bacteria community was beneficial as only 58% of germ-free larvae survived to adults while even less immune-compromised germ-free larvae (22%) developed into adults following infection with C. albicans (Glittenberg et al., 2011b). Despite C. albicans being restricted to the gut, it still caused systemic infection as detected by drs expression from the fat body.
Using a similar feeding protocol, we established a GI infection model for C. glabrata in Drosophila larvae. We screened (one at a time) a library of deletion mutants including 196 C. glabrata transcription factors (Schwarzmüller et al., 2014) and we identified five that did not grow or activate systemic immunity in larvae. From these five fungal mutants, C. glabrata deficient for ADA2 was highly sensitive to very low concentrations of H 2 O 2 (a hallmark of oxidative stress) ex vivo. Inside the host, the ada2 mutant was only able to grow in ROSsuppressed Drosophila while wild type C. glabrata with an additional copy of ada2 was able to grow and kill wild type larvae. Our results suggest that ADA2 mediates C. glabrata defenses against ROS and is therefore a potential target to diminish C. glabrata growth during GI infection.

Infection experiments
We followed a modified gastrointestinal infection model (Glittenberg et al., 2011b) based on a previously established bacterial oral infection protocol (Basset et al., 2000). Accordingly, every C. glabrata mutant strain was tested one-by-one. Prior to infection, 10mL of yeast extract peptone dextrose (YPD) (Y1375, Sigma-Aldrich, USA) was inoculated with a single colony of yeast and incubated on a rotating shaker at 30°at 120 rpm for 16 hr. The inoculum was pelleted by centrifuging at 4°at 4000 rpm for 4 min and the supernatant was discarded. The pellet was washed in 10mL of 1X PBS and centrifuged again at 4°at 4000 rpm for 4 min and the supernatant was discarded. A single banana was homogenized and 450mL was added to 2mL eppendorf tubes. To this, 250mL of pathogen (OD = 200) was also added. As a control, banana without pathogen was used. Five-day (third instar) larvae were collected by washing from the fly food and caught in a sieve. They were starved for one hour before infection. $100 larvae were added to each eppendorf tube, plugged with breathable foam with space for fermentation and allowed to feed for 30 min. The mixture then was transferred to a standard fly medium and incubated at 30°. Lethality of larvae following infection was determined by counting surviving pupae. 48 hr after infection.

TFKOs mutants screen
The transcription factor knockout (TFKO) library screen was conducted following the above oral infection protocol in w 1118 ; drs-GFP 5-day old larvae. Two days after infection, pupae were visualized under UV light to confirm that they were unable to activate immunity. Further to this, drs-GFP were infected with the complement strain of each hit to re-confirm that when the knocked-out gene was re-inserted, immunity would be activated.

Yeast counts from larvae by counting CFUs
Following infection, larvae were washed in 100% ethanol, rinsed in sterile water and transferred to normal fly food. Larvae were homogenized (at indicated time points) in 200 mL YPD media (Y1375, Sigma-Aldrich, USA). Serial dilutions were made and 50 mL, plated on YPD plates (Y1500, Sigma-Aldrich, USA). Plates were incubated at 30o and colonies were measured accordingly.

Gene expression analysis
Drosomycin expression was determined from five independent biological samples consisting of 5 infected third instar larvae following infection. Samples were compared to banana-fed flies of the corresponding time point. Larvae were collected at the desired time points and washed in 100% ethanol and sterile water. RNA was extracted using Purification Plus Kit (48400, Norgen -Biotek, Canada) and cDNA was prepared from 0.5 mg total RNA using Maxima First Strand cDNA Syntesis Kit (K1672, Thermo Scientific, UK). Triplicate cDNA samples were amplified with the SensiFASAT SYBR No-ROX Kit (BIO-98020, Bioline, UK) in a Corbet Rotor-Gene 6000 QPCR machine (Qiagen, UK) according to the manufacturer's protocols. Expression values were calculated using the DDCt method and normalized to rp49 expression levels (Schmittgen and Livak 2008).
Measuring production of reactive oxygen species (ROS) The amount of ROS was measured by quantifying the amount of hydrogen peroxide produced per larvae immediately after oral infection. The samples included three independent biological samples of 10 third instar larvae from non-infected, banana-fed and infected with C. glabrata samples. The amount of hydrogen peroxide produced from the sample was determined using Amplex red hydrogen peroxide/ peroxidase assay kit (A22188, Invitrogen, USA) following manufacturer's instructions.
C. glabrata deletion mutant construction Construction of C. glabrata transcription factor deletion mutants used in this study was previously described (Schwarzmüller et al., 2014).
C. glabrata complementation and overexpression strain construction C. glabrata ORFs previously cloned into the pDONR221 entry vector using GATEWAY cloning technology (Thorne et al., 2011) were shuttled into a pAG423GPD-ccdB destination vector (AddGene) using LR clonase. Destination vectors carrying the C. glabrata ORF were transformed by electroporation into the corresponding C. glabrata deletion mutant for complementation or into the parental C. glabrata Dhis3 strain for overexpression. Empty pAG423GPD-ccdB vector was transformed into C. glabrata as a control. Correct transformants were selected for growth on SC -his media. Three independent transformants of each strain were collected.

Phenotypic screening
Overnight cultures of C. glabrata strains were normalized to OD600 0.1 in sterile water. The normalized suspensions were aliquoted into a 96-well plate and diluted ten-fold across six wells to create serial dilutions. Using a multichannel pipette, 5 ml aliquots of the serial dilutions were spotted onto SC agar plates supplemented with stress-inducing chemical agents (Table S2). Plates were incubated at 30°, unless otherwise stated, and imaged daily. Upon visual inspection, phenotypes for each mutant were scored relative to the parental strain into six categories: mild sensitivity (MS), sensitive (S), severe sensitivity (SS), no growth (NG), improved growth (IG) or no phenotype. Three biological replicates were performed for all phenotypic screens.

Growth analysis
Overnight cultures of C. glabrata were normalized to OD600 0.1 in fresh YPD and distributed into a flat-bottomed 96-well plate in 100 ml aliquots. Optical density readings were taken every 10 min in a VersaMax Absorbance Microplate Reader set to 30°with shaking between reads. Growth rate was measured by calculating the doubling time for each mutant during exponential phase (OD600 0.4 -0.8).

Data availability
Drosophila and Candida strains are available upon request. All data pertaining to this study can be found in the figures and supplemental material. A list of the C. glabrata mutants used for the primary screen can be found in table S3 of Schwarzmüller et al. 2014.

RESULTS
The host response following C. glabrata GI infection To establish an oral infection screen for a C. glabrata mutant library, we developed a third instar larval yeast-feeding protocol (see material and methods for details). Briefly, we fed larvae a C. albicans or S. cerevisae or C. glabrata/banana mixture while in parallel to each infection, larvae fed with banana only, served as our control. For these experiments we made use of third instar w 1118 larvae (a wild type strain and the genetic background of the mutant fly strains used in this study). First, we determined the dynamics of host immune triggering after GI infection with reference yeast strains compared to the banana-only control ( Figures 1A-C). As a measure of Toll pathway activity we monitored drs mRNA levels at specific time-points ( Figure 1A-C). On average, 24h was the time point where the most elevated drs gene expression levels was seen ( Figures 1A-C). However, there was a significant difference in the capacity of the three fungi to activate drs gene expression with C. albicans-induced activation almost 10x more compared to S. cerevisiae and C. glabrata ( Figure 1A-C). Nevertheless, compared to non-infected larvae ( Figure 1D), there was a robust fat body induction of a drs-GFP marker in larvae infected with the reference C. glabrata strain Cg2001H ( Figure 1E). This confirmed a systemic response after feeding and was reminiscent of the same result in C. albicans infection [37]. In addition, there was significant increase in ROS production in Cg2001H-infected larvae over and above the banana control ( Figure 1F). Finally there was a significantly reduced survival of GI infected larvae at 24h compared to both the banana control and to non-infected larvae ( Figure 1G). Of note, compromised survival of w 1118 was not an effect of that specific genetic background as it was also observed in Oregon R and w 1118 ; drs-GFP larvae ( Figure S1, A and B in File S1).
Exploring individual strain infectivity in a library of C. glabrata deletion mutants After determining that the host responses to GI infection following C. albicans and C. glabrata infection were comparable (albeit not of the same intensity), we used the system to screen a library of C. glabrata deletion mutants (TFKO library) for their ability to activate the Toll pathway. The library included deletions of 196 C. glabrata transcription factors (TF), with well-characterized roles in other yeast species (Schwarzmüller et al., 2014). In this manner, we would be able to pinpoint to gene regulatory networks organized by specific transcription factors that are involved in infectivity. Table S1 presents all the strains used for the screen. We used w 1118 ; drs-GFP as a proxy for immune induction and virulence. Five of the 196 C. glabrata TF mutants (Δada2, Δbas1, Δhir3, Δino2, Δmet31) did not activate drs expression at 24h following infection (Figure 2A). To ascertain that this phenomenon was due to the deleted TFs, complementation strains were constructed. Complemented strains were able to activate drs to a degree that was statistically indistinguishable from the C. glabrata  Figure 2B). In contrast, this was not the case when the TFKO strains were complemented with just the empty vector ( Figure S2A in File S1; for a list of the complementation strains used see Table S1). We used Colony Forming Units (CFUs) to monitor the presence of Cg2001H inside GI infected larvae. W 1118 ; drs-GFP larvae presented similar kinetics with w 1118 larvae ( Figure 2C and Figure 2D respectively), with Cg2001 cleared at approx. 64h post-feeding. In contrast, immune-deficient w 1118 ; dif-key larvae showed delayed clearance with a considerable CFUs of Cg2001H still at 96h post-feeding ( Figure  2E). Direct comparison between w 1118 and w 1118 ; dif-key larvae showed that at 48h post-infection, there was a significant difference in CFUs with a 24h difference in clearance ( Figure 2F). Nevertheless, the five TFKO strains that did not activate drs in our screen were cleared significantly faster than Cg2001H with zero or much lower CFU count at 24h post-infection ( Figure 2G).
Ex vivo phenotypic characterization of C. glabrata TF mutants failing to grow inside the host To pinpoint a mechanism by which the C. glabrata mutants may be cleared in the larval host, the mutants were tested for growth on several conditions relating to the host environment. This included nutrient limitation tolerance, pH and temperature sensitivity and susceptibility to antifungals ( Figure 3A; for a list of all ex vivo conditions tested see Table S2). C. glabrata Δbas1 and Δhir3 grew similarly to the parental  strain Cg2001 under all stress conditions tested while Δino2 was susceptible to 0.01%SDS ( Figure 3A). In addition, Δmet31 was susceptible to H 2 O 2 , 20mM HU and 0.01% SDS ( Figure 3A). The most susceptible phenotypes were observed for C. glabrata Δada2 which was susceptible to growth on 10 out of 13 of the tested conditions relating to growth under decreased temperatures, nutrient limitation, oxidative stress, cell membrane stress, cell wall stress and antifungal drug tolerance ( Figure 3B, Table S3).
The five TF deletion mutants were further investigated for any possible dysfunctions, which may relate to their reduced ability to persist in the larval host. Initially, we monitored growth under non-stressed conditions at 30°. Outside the host, all mutant strains displayed growth curves comparable to the reference strain ( Figure S3A in File S1). These growth rate measurements indicated that these TF mutants were not growthdefective. However, C. glabrata Dada2 and Dbas1 deletion mutants displayed an increased (longer) generation time ( Figure S3B in File S1).
The susceptibility of C. glabrata Δada2 to oxidative stress inducing agents H 2 O 2 and menadione were of particular interest since increased levels of ROS were recorded upon Cg2001H infection in earlier experiments ( Figure 1F). Therefore, we explored the hypothesis that ADA2 was indispensable for resistance to toxic levels of oxidative stress, something that would be the case when encountering the localized epithelial immune response of the Drosophila GI tract.
The role of the C. glabrata TF ADA2 in Drosophila GI infection The Drosophila dual oxidase (dDuox) has been shown to regulate ROS in the fly intestine (Ha et al. 2005). When dDuox was knocked down in enterocytes (w 1118 ; np1-GAL4; Duox RNAi ), larval survival following GI infection with Cg2001H was significantly reduced ( Figure 4A). This showed that absence of ROS production increased susceptibility of larvae to GI infection. In contrast, survival of these larvae following infection by the Dada2 deletion mutant was significantly improved compared to Cg2001H ( Figure 4A). This showed that when both host ROS production as well as pathogen ROS defenses were absent larval survival was largely restored. This was not the case when Dada2 was complemented with a wild type copy of the ada2 gene ( Figure 4A). Consistent with the role of ada2 in the process, complementation of the Dada2 mutant with an empty vector did not compromise larval survival ( Figure 4A). Finally, adding an extra copy of the ada2 gene in the Cg2001H reference strain made the latter more virulent than normal when host ROS production was supressed ( Figure 4A). This underlined the role of ada2 as a major regulator of C. glabrata virulence beyond ROS defenses in the context of gastrointestinal infection. In contrast, none of the other mutants compromised larval survival when overexpressed ( Figure 4B).
( Figure S4, A and B in File S1 respectively), w 1118 (as the genetic background of all of the mutants used; Figure S4C in File S1) and Oregon R flies (as an extra genetic background; Figure S4D in File S1). Overexpression strains of the rest of the hits were also constructed (Table S1) and their pathogenicity was examined in the same host background like the ada2 overexpression strain. Survival of larvae showed that hyper virulence of ada2 overexpression was unique among the other TFKO hits of the original screen ( Figure 4C). CFU measurements revealed that there was an expansion of both Cg2001H and Cg2001H overexpressing ada2 (Cg2001H::pADA2) when Duox gene expression was knocked-down ( Figure 4C) resulting in increased larvae lethality. However, in case of infection of a wild-type host strain (with normal ROS production) there was a significant increase in CFUS of Cg2001H::pADA2 in contrast to the clearance of the reference Cg2001H strain ( Figure 2C and Figure 4D). This meant that an extra copy of ada2 allowed increased growth of the pathogen. Moreover, as seen by the larval survival assays, elevated growth of Cg2001H::pADA2 compromised the host significantly more than the wild type C. glabrata. However, this growth advantage was lost when larvae were unable to produce ROS. Taken together with the inability of ada2 to grow in even the lowest of H 2 O 2 concentrations, the above results suggest that the TF ada2 is important for regulating C. glabrata defenses against host-generated ROS.

DISCUSSION
C. glabrata is becoming an important problem in persistent hospital infections (Pfaller and Diekema 2007). Starting as a commensal microbe it can generate systemic challenges that have an increasingly incurable outlook (Kaur et al., 2007). The growing resistance of C. glabrata to various antifungals including azoles underlines the above and indicates the need to develop new drugs and/or therapies to target this pathogen (Perlin 2007;Pfaller et al., 2010;Perlin 2014;Singh-Babak et al. 2012). To this end, we screened a C. glabrata mutant library of transcription factor and co-factor deletions (TFKO library, see Schwarzmüller et al., 2014) to identify possible novel virulence factors responsible for C. glabrata pathogenicity in the GI tract. Five candidate genes were identified and their ability to infect the host was explored further. Five deletion strains showed accelerated clearance by the host gut. Considering that gut conditions are a new stress environment for the pathogen we followed a series of ex vivo stress response tests including oxidative stress sensitivity, antifungal sensitivity, pH and temperature sensitivity. Between the five deletion mutant hits, Δada2 deletion showed severe growing limitations under oxidative stress response tests. Therefore, this particular mutant was studied further. In addition, ada2 deficient C. glabrata was unable to activate a robust systemic immune response. This was either due to the fact that the mutation altered the cell wall in a manner rendered unrecognizable by GNBP3 or because proteases normally secreted and able to activate PSH directly were not upregulated. More work on the re-wiring of transcription of wild type C. glabrata and ada2 mutants within the host is needed to distinguish between these two possibilities.
The side of the host The synthesis of ROS (including H 2 O 2 , hydroxyl radicals and superoxide) as an immune defense is the first line immune response by phagocytes upon encountering a fungal infection in humans. Similarly, the production of both ROS and AMPs are important features of Drosophila epithelial immunity (Tzou et al., 2000;Ha et al., 2005;Buchon et al., 2009). ROS can cause damage to DNA, RNA, and proteins and trigger the oxidative degradation of lipids in cell membranes. GI infection of Drosophila larvae by C. glabrata increased ROS levels significantly when compared to non-infected or banana-only treated controls. When the percentage of larvae surviving to pupae was determined, we found that although feeding with banana activated a low level ROS induction ( Figure 1F) this did not affect the survival of the organism. However, infection with the reference strain Cg2001H did impact survival with approx. 60-65% of larvae developing to pupae. Oral infection with C. glabrata activated epithelial immunity in the GI tract by inducing both ROS production as well as leading to a systemic activation of the Toll pathway. Lack of the Toll and Imd pathways delayed but did not stop the clearance of the pathogen, implying a major role for ROS in the equilibrium between host and pathogen. The crucial role of ROS was further supported when we recorded significantly increased levels of host lethality in larvae that were unable to produce ROS.
The side of the pathogen Five deletion mutants from the TFKO library showed accelerated clearance and inability to induce an immune response inside the host. Growth curves of the 5 deletion mutants showed that all the mutants were able to successfully grow at 30°in YPD media, outside of the host. Moreover, the generation times for the 5 deletion mutants were similar, with Δada2 and Δbas1 displaying a marginally shorter generation time than the other strains. Nonetheless, all the deletion strains did not display any detectable growth defects that would be the cause of its rapid elimination from the host. The ability of the deletion mutants to grow under different stress conditions included tests for nutrient limitation tolerance, sensitivity to antifungals, defects in cell wall/membrane and sensitivity to oxidative stress conditions, for example. All but one strain were able to grow under all the conditions, albeit reduced and limited growth was observed. However, Δada2 deletion mutant showed high sensitivity to the oxidative stress conditions with very restricted or no growth. As previously mentioned, wild type C. glabrata has shown to grow in extremely high H 2 O 2 conditions (Cuéllar-Cruz et al., 2008). Thus, the inability of Δada2 mutants to grow in low H 2 O 2 conditions indicated that the ada2 gene is essential for its resistance to the ROSmediated immune defense in the Drosophila gastrointestinal tract. To this direction, Δada2 mutant was investigated further using ROS-suppressed flies.

Δada2 mutant and the interaction with hostgenerated ROS
The open reading frame (ORF) of ada2 in C. glabrata is CAGL0K06193g and is known to be involved in the transcriptional activation of RNA polymerase II. Moreover, ortholog(s) are also implicated in chromatin binding and histone acetyltransferase. This suggests that the gene function of ada2 is very broad and is likely to be involved in the activation of various proteins. Indeed, genomewide mapping of ADA complexes in Candida albicans have shown that ADA2 is recruited to 200 promoters of genes involved in different stress-response functions (including oxidative resistance) and metabolic responses (Sellam et al. 2009). Consistent with these results, our ex vivo experiments clearly showed that Δada2 deletion mutants were highly sensitive to various oxidative stress conditions suggesting that among the many functions of this gene is also the co-ordination of oxidative stress coming from the ROS-producing host immune reaction. In C. albicans, host-derived ROS affects only fungal cells exposed to phagocytosis and epithelial immunity rather than those in systemic infection (Enjalbert et al., 2007). Mechanisms to avoid or defend against host-derived ROS in this fungus include extracellular anti-oxidant enzymes of the super dismutase (SOD) family of enzymes (Frohner et al., 2009;Hwang et al., 2002) all involving SAGA/ADA-mediated co-activation (Sellam et al. 2009). Both its relation with ROS stress resistance and the fact that in C. albicans ADA2p binds the promoters of several drug transporters (mdr1, cdr1, cdr4, qdr1, ycf1, flu1 to name a few) makes ADA2 an important target for the development of new antifungals if one wanted to eliminate C. glabrata from the GI tract (Sellam et al., 2009) (see below).
When orally infected with Δada2 mutants, larvae with supressed ROS production were able to survive like wild type controls, indicating that the inability of the gut to synthesize ROS had no impact on host surviving under these conditions. However, infection with the wild type C. glabrata Cg2001H strain resulted in a significant decrease in survival of the same larvae. This indicated that in the absence of the ada2 gene, ROS-suppressed larvae were able to survive better. When ada2 gene was re-introduced, virulence was restored. Moreover, when this gene was overexpressed, survival of ROS suppressed flies decreased significantly when compared to infections with the Cg2001H strain. Finally, in wild type (w 1118 ) larvae with functioning Duox, ada2 overexpression strain was able to persist and expand three hours after infection while the fungal load of wild type C. glabrata decreased. This indicated that an additional copy of the ada2 gene made this strain hyper-virulent, significantly reducing the survival of the host and enabling expansion within the host. Contrastingly, survival of w 1118 ; drsGFP larvae when infected with the remaining 4 overexpression strains of the deletion mutant hits (Δbas1, Δhir3, Δino2 and Δmet31) was similar to wild type C. glabrata highlighting the specific effect of ada2 overexpression.
The induction of ROS has been shown to be a rapid and important immune defense following oral infection in Drosophila (Ha et al., 2005). Flies that lack the ability to generate ROS upon natural infection succumb to infection. Ha and colleagues (Ha et al., 2005) demonstrated ubiquitous expression of Duox-RNAi resulted in increased mortality following natural infection with Ecc15. Likewise, when Duox-RNAi was restricted to the gut, flies displayed a similar level of mortality. Interestingly, when Duox-RNAi was introduced to the main immune tissues in systemic immunity in Drosophila (the fat body/hemocytes), survival of flies was unaffected (Ha et al., 2005). This indicates that ROS plays a major role in the host resistance during GI infection. Therefore, the necessity for ADA2-mediated transcription to resist host-generated ROS as implied by our results may be a gut-specific interaction.
Indeed, a recent study has indicated that although an independently constructed ada2 mutant exhibited susceptibility to three classes of antifungal drugs (i.e., azoles, echinocandins, and polyenes) as well as cell-wall perturbing agents it was hyper-virulent in a mouse model of systemic infection (Yu et al., 2018). This was contrary to what we have observed in our Drosophila model. However, since the route of infection was different our hypothesis is that GI vs. systemic immunity may influence the virulence outcome of ada2 mutants. More work is needed to determine the difference in the interaction between ada2 mutants with the various aspects of host defense.

ACKNOWLEDGMENTS
We thank the members of the Ligoxygakis and Haynes labs for discussions, Mario Iurlaro for initial experiments and Peter Burns for help with fly breeding. This work was funded by BBSRC grant BB/ F005210/1 (to KH) and NC3Rs grant NC/K500392/1 (to PL and RM). IK was supported by a postdoctoral fellowship from the Bodossaki Foundation. We would like to dedicate this paper to our dear friend and colleague Prof Ken Haynes, who passed away after the early on line publication of this work following a long and brave battle with cancer. Our community suffered a great loss but his memory lives on. We will always remember him as someone with a big heart and a sharp intellect, a great mentor who looked out for junior colleagues and students.