The Fungal Histone Acetyl Transferase Gcn5 Controls Virulence of the Human Pathogen Candida albicans through Multiple Pathways

Fungal virulence is regulated by a tight interplay of transcriptional control and chromatin remodelling. Despite compelling evidence that lysine acetylation modulates virulence of pathogenic fungi such as Candida albicans, the underlying mechanisms have remained largely unexplored. We report here that Gcn5, a paradigm lysyl-acetyl transferase (KAT) modifying both histone and non-histone targets, controls fungal morphogenesis – a key virulence factor of C. albicans. Our data show that genetic removal of GCN5 abrogates fungal virulence in mice, suggesting strongly diminished fungal fitness in vivo. This may at least in part arise from increased susceptibility to killing by macrophages, as well as by other phagocytes such as neutrophils or monocytes. Loss of GCN5 also causes hypersensitivity to the fungicidal drug caspofungin. Caspofungin hypersusceptibility requires the master regulator Efg1, working in concert with Gcn5. Moreover, Gcn5 regulates multiple independent pathways, including adhesion, cell wall-mediated MAP kinase signaling, hypersensitivity to host-derived oxidative stress, and regulation of the Fks1 glucan synthase, all of which play critical roles in virulence and antifungal susceptibility. Hence, Gcn5 regulates fungal virulence through multiple mechanisms, suggesting that specific inhibition of Gcn5 could offer new therapeutic strategies to combat invasive fungal infections.

www.nature.com/scientificreports www.nature.com/scientificreports/ GCN5 deletion alters cell surface components and FKS expression. We next analysed in more detail the effect of Gcn5 loss on cell wall architecture by quantifying all cell wall components (chitin, β1,3-glucan, branched β1,6-glucan and mannoproteins) using a FACS-based triple staining assay 60 . Both β-D-glucan and chitin levels were about 2-fold elevated in gcn5∆/∆ cells when compared to the wild type control or the restored strain. Outer mannoprotein levels were unchanged, indicating that gcn5∆/∆ causes specific cell wall defects rather than gross effects (Fig. 2C). We hypothesized that the increased β-D-glucan content in the gcn5∆/∆ mutant may be due to increased expression of glucan synthases such as FKS, which synthesises β-1, 3-D-glucan. To test this, we quantified mRNA expression levels of all three FKS genes, FKS1, FKS2 and FKS3, encoded in the C. albicans genome. Unexpectedly, we found that expression of the CSP target Fks1 was significantly reduced upon removal of Gcn5, whereas both FKS2 and FKS3 were significantly upregulated (Fig. 2D). This implies a compensatory inverse regulation of FKS genes caused by the loss of Gcn5. Thus, we believe that the increased expression of FKS1 upon CSP treatment in wt cells explains the CSP tolerance, which in turn is impaired when Gcn5 is ablated. In other words, the FKS2/3-triggered compensatory increase in glucan is not sufficient to restore CSP tolerance when GCN5 is deleted. We cannot exclude a distinct glucan distribution or an aberrant integration of newly synthesized glucan into the functional cell wall architecture. Logarithmically growing cells of SC5314 wild-type (wt), homozygous deletion (gcn5Δ/Δ) and restored (gcn5Δ/Δ::GCN5) Candida strains were plated on YPD plates supplemented with 10% fetal calf serum. Colony morphology was analysed after incubating for 3 days at 37 °C. Photographs were taken using a Discovery V12 Stereoscope equipped with an Axiocam MR5 camera (Zeiss). Scale bar = 1 mm. (B) Representative confocal microscopy image showing fungal cells stained with calcofluor white (CFW) to detect cell wall chitin. Cells were fixed in 4% p-formaldehyde for 2 hours, washed and stained with CFW 1 mg/ml for 5 min. Differential Interference Contrast (DIC) and UV light images (UV) of the same cells are shown at 60x magnification. Scale bar = 5 µM. (C). Representative confocal DIC images showing the hyphal morphology of strains at indicated temperatures. Logarithmically growing cells in YPD supplemented with 10% FCS were fixed in 4% p-formaldehyde for 2 hours and washed and images were taken with an LSM 700 Zeiss Confocal microscope at 60x magnification. Scale bar = 5 µM. (2019) 9:9445 | https://doi.org/10.1038/s41598-019-45817-5 www.nature.com/scientificreports www.nature.com/scientificreports/ To test whether loss of Gcn5 alters cell membrane permeability, which may explain the altered drug sensitivity, we measured the kinetics of fluorescein diacetate (FDA) uptake, whose linear uptake is solely a function of lipid membrane permeability 61 and thus independent of transporters ( Supplementary Fig. S2C). Indeed, CSP increased FDA uptake in the wild type and restored strains, whereas basal FDA uptake in gcn5∆/∆ cells was already higher, indicating altered membrane lipid permeability that was not further increased by CSP (Supplementary Fig. S2C). To conclude, these results show that genetic ablation of GCN5 alters the cell wall composition, FKS expression, and membrane lipid permeability. transcriptional response of wild type and gcn5Δ/Δ cells upon caspofungin stress. To dissect the Gcn5-dependent regulatory networks that may be underlying the observed phenotypes, we performed RNA-seq of gcn5∆/∆ cells and wild type in the presence and absence of CSP (Fig. 3). Growth curves and CSP sensitivity in liquid culture suggested a concentration of 10 ng/ml as appropriate, since this concentration exerted . Transcriptional response of wt and gcn5Δ/Δ cells upon caspofungin stress. Total RNA was isolated from cultures of SC5314 wild-type (wt) and homozygous deletion strain (gcn5Δ/Δ) grown in the absence or presence of caspofungin and subjected to RNA-seq analysis. The bioinformatics pipeline for data analysis and hierarchial clustering is described in materials and methods. (A) Genetic removal of GCN5 increases caspofungin susceptibility. Growth of indicated strains in YPD containing or lacking 10 ng/ml caspofungin (CSP10) at 30 °C. Data represent the means from three independent growth analyses (±SEM). Solid and dotted lines indicate untreated and caspofungin-treated cultures, respectively. (B) Heat map of hierarchical clustered and differentially expressed genes in wt and gcn5Δ/Δ cells in response to caspofungin treatment (CSP). Lanes 1 and 2 compare 15 min and 45 min CSP-treated wt cultures with untreated YPD-grown cells, respectively. www.nature.com/scientificreports www.nature.com/scientificreports/ significant toxicity but still allowed for growth of all cells (Fig. 3A). The gcn5∆/∆ mutant showed sensitivity to 10 ng/ml CSP but retained more than 85% viability after 45 min but not at 60 min, where 30% cells were inviable ( Fig. 3A; data not shown). Hence, we considered the time points 15-and 45-min post CSP treatment as optimal for RNA-seq. Differentially regulated genes were included when expression experienced a ≥ or ≤log 2 1.5-fold change, with an adjusted P-value cut-off ≤ 0.05 (Fig. 3, Supplementary Table S4).
We first compared transcript levels of YPD-grown gcn5∆/∆ to YPD-grown wild type cells and found 336 differentially regulated genes (Fig. 3B, lane 5). Out of these, 235 genes were induced, and 101 genes were repressed (Fig. 3C). The distribution of genes responding to GCN5 deletion classified according to gene ontology (GO)-slim biological processes included major-enriched (cell adhesion, oxidation-reduction, transmembrane transport and pathogenesis) and minor-enriched (arginine biosynthesis, ergosterol and hydrogen sulphide biosynthesis, heme transport and zinc sequestration) as depicted in Supplementary Fig. S3A. Notably, genes implicated in pathogenesis were exclusively induced, except for the alpha-1,3-mannosyltransferase MNN14. Three-quarters of the differentially regulated genes were up-regulated in gcn5∆/∆ cells, implying that Gcn5 target genes are primarily negatively regulated. Together, our genome-wide analysis of gcn5∆/∆ cells suggests that GCN5 deletion changes the landscape of several C. albicans transcriptional networks by regulating various target gene sets.
To determine whether alterations in the transcriptome potentially underlies the increased antifungal susceptibility of gcn5∆/∆ cells, we compared the effect of 15 and 45 min CSP treatment on the global transcriptomes of gcn5∆/∆ cells versus wild type cells (Fig. 3B, lanes 6 and 7). The Venn diagrams include the genes regulated at all time points from 0 min (untreated),15 min to 45 min CSP-treated in gcn5∆/∆ compared to wild type cells (Fig. 3C). We found that 15 min and 45 min CSP treatment showed approximately 70% and 50% overlap of induced (135 genes) and repressed (46 genes) genes, respectively (Fig. 3C). We also combined both 15 min and 45 min CSP-treated datasets and found that 154 genes were induced by CSP in both the gcn5∆/∆ and wild type cells. Some 81 genes were induced specifically in the absence of GCN5, while 115 genes were induced specifically upon CSP stress (Fig. 3C, upper panel). Surprisingly, given the differential sensitivities of the wild type and gcn5∆/∆ mutant to CSP, the CSP-mediated transcriptional response in wild type (Fig. 3B, lanes 1 and 2) and gcn5∆/∆ mutants (Fig. 3B, lanes 3 and 4) after 15 min and 45 min) revealed a striking overlap (Fig. 3B, lanes 1-4). The major biological processes differentially affected by the loss of Gcn5 included cell adhesion, oxidation-reduction, transmembrane transport, pathogenesis, lipid catabolism and iron homeostasis ( Supplementary Fig. S3B). The inability of gcn5∆/∆ cells to fully induce the normal CSP response explains the increased antifungal susceptibility.

Lack of GCN5 affects oxidation-reduction, transport, pathogenesis and ion homeostasis. To
validate key predictions of the transcriptomics datasets, we employed quantitative real-time RT-PCR on hallmark genes known to be implicated in biological processes putatively affected by loss of Gcn5 ( Supplementary  Fig. S3A,B). Expression of at least three genes in each biological process was validated. The superoxide dismutase genes SOD4 and SOD5, and the ferric reductase CFL2 were previously implicated in oxidative stress resistance, as well as in the yeast-to-hyphae transition. In addition, SOD5 has been linked to CSP-response [62][63][64] . Indeed, RT-qPCR analysis revealed approximately 4-fold, 3-fold and 29-fold increased SOD4, SOD5 and CFL2 mRNA levels, respectively, in gcn5∆/∆ cells when compared to the wild type (Fig. 4A). However, lack of GCN5, significantly elevated CFL2 levels upon short-term treatment (CSP-15′), with a reduction at CSP-45′. As expected, CSP induced SOD5 expression in wild type cells in a time-dependent manner (Fig. 4A), whereas, gcn5∆/∆ mutants were completely unable to induce SOD5 upon CSP challenge.
Next, we checked the expression levels of HYR1, which encodes a GPI-anchored cell wall protein involved in hyphal regulation through the MAPK signaling genes CEK1 and CEK2. These genes are required for the pathogenesis and cell wall-mediated invasion and pheromone responsive MAPK cascade, respectively 65,66 . Loss of GCN5 induced approximately 4-fold, 1.5-fold and 8-fold of HYR1, CEK1 and CEK2 mRNAs, respectively. Prolonged CSP treatment slightly increased CEK1 gene expression, contrary to the expression of HYR1 and CEK2 genes that were slightly repressed. 45 min CSP treatment induced CEK1 expression by about ~3-fold (Fig. 4B). These results indicate that Gcn5 negatively regulates Cek1/Cek2 and Hyr1. Additional hyphal regulators include EFG1, NRG1, TUP1, TEC1, and hyphae-associated genes showing a predominant expression in filaments include HWP1, ECE1 and RBT5 29,31,67 . Of note, these genes are also involved in biofilm formation and some are essential for fungal virulence 31,66-69 . Since, GCN5 strongly affected hyphal formation (Fig. 1A,C), we tested whether additional hyphal regulators are also regulated by GCN5. Ablation of GCN5 showed a ~5-fold and 2-fold basal upregulation of HWP1 and ECE1 transcripts, respectively. In contrast, TEC1 mRNA was reduced ~2-fold when compared to wild type cells. Furthermore, expression of the repressor genes NRG1 and TUP1 were unaffected by the lack of Gcn5 ( Supplementary Fig. S4).
Next, we quantified expression of the zinc permease Zrt2 and genes encoding multidrug resistance transporters such as the Mdr1 major facilitator and the ATP-binding cassette transporter Cdr4. In wild type cells, CSP induced MDR1 expression, but significantly reduced expression of CDR4 and ZRT2. Lack of GCN5 rendered cells unable to drive CSP-responsive MDR1 expression and reduced basal levels of CDR4 and ZRT2 (Fig. 4C). In addition, we quantified expression of iron homeostasis genes. The high-affinity iron permease genes FTR1 and FTR2 were differentially regulated upon loss of GCN5 and CSP treatment, with 10-fold up and 6-fold down-regulated mRNAs in gcn5∆/∆ cells, respectively. Notably, expression of FTR1 and FTR2 encoding iron permeases was compromised upon CSP treatment in gcn5∆/∆ cells but induced in wild type in a time-dependent manner (Fig. 4D). Interestingly, gcn5∆/∆ cells showed 13-fold higher basal expression of the cell wall protein RBT5, with ~3-fold further upregulation upon CSP treatment (CSP-15′), whereas prolonged exposure significantly diminished RBT5 expression (Fig. 4D). Taken together, removal of GCN5 differentially regulates genes involved in oxido-reduction, ABC and MFS drug transporters, filamentation regulators, MAPK signaling modules, as well as iron / zinc homeostasis and heme acquisition, which requires Rbt5 and affects cell wall architecture by regulating its proteome including adhesins 70-74 . www.nature.com/scientificreports www.nature.com/scientificreports/ GCN5 regulates cell adhesion through Efg1. The bHLH transcription factor Efg1 integrates signals through upstream cAMP-PKA pathways 31,35,75 to control morphogenesis, CSP tolerance and adhesion 76 . Thus, we asked whether lack of GCN5 affects Efg1 expression. We noticed a significant reduction in EFG1 mRNA in gcn5∆/∆ www.nature.com/scientificreports www.nature.com/scientificreports/ cells when compared to wild type cells, which was not seen in restored strains (Fig. 5A), and confirmed the RNA-seq data (Fig. 3). Since Efg1 strongly impacted adhesion, we also quantified adherence properties by testing binding to polystyrene-coated plastic surfaces using the crystal violet readout. Indeed, cells lacking GCN5 showed drastically attenuated adherence to polystyrene surfaces, which was not seen in the restored strain (Fig. 5B).
Efg1-mediated adhesion primarily engages the ALS agglutinin adhesion family 76 . Given that loss of GCN5 increases susceptibility to CSP and decreases Efg1-mediated cell adhesion family, we tested the effect of CSP on Als1 expression (Fig. 5C). Wild type cells dramatically elevated Als1 levels upon CSP treatment, but the upregulation of ALS1 was completely abolished in cells lacking Gcn5 (Fig. 5C). To reveal a potential role of other adhesins, we tested basal expression of all GPI-anchored ALS genes encoded in the Candida genome in wild type and gcn5∆/∆ mutants ( Supplementary Fig. S5). Interestingly, like ALS1, ALS2/3/4 were downregulated in gcn5∆/∆ cells, whereas ALS5/6/7/9 mRNAs were significantly upregulated ( Supplementary Fig. S5). Although the function of most ALS genes is unknown, the Gcn5-dependent dysregulation of well-established adhesins Als1 and Als3 explains the adhesion defects (Fig. 5C, Supplementary Fig. S5).
CSP inhibits β-1,3-D-glucan deposition by Fks1, whose overexpression confers resistance to CSP 77 . Wild type cells displayed significantly increased FKS1 expression after 15 and 45 min CSP treatment. Conversely, FKS1 mRNA was reduced in gcn5∆/∆ cells, and CSP treatment no longer increased FKS1 mRNA levels ( Fig. 5D), whereas both FKS2 and FKS3 were significantly upregulated (Fig. 2D). These data explain the CSP hyper-susceptibility and suggest a pivotal role of Gcn5 in regulating glucan deposition and cell wall function, including a coordinated action in concert with EFG1-mediated adhesion.
Gcn5 is required for intracellular survival in BMDMs and essential for fungal virulence. Several phenotypes adopted by gcn5∆/∆ cells implied that Gcn5 may be involved in fungal virulence, as suggested by a previous report 56 . Thus, we investigated the role of GCN5 in virulence in vivo using mouse models, and quantified the susceptibility of fungal cells to killing by innate immune cells. We used a murine Candia infection model and injected 1 × 10 5 Candida cells through the lateral tail vein 78 . Mice infected with gcn5∆/∆ cells or mice receiving only PBS showed no weight loss. By contrast, mice receiving wild type or restored Candida strains lost approximately 20% of their weight over seven days (Fig. 6A). Remarkably, wild type and GCN5 reconstituted cells efficiently killed animals, whereas gcn5∆/∆ cells were avirulent with all mice being able to clear the systemic infections (Fig. 6B). www.nature.com/scientificreports www.nature.com/scientificreports/ To uncover a possible cellular mechanism, we challenged primary BMDMs with fungal strains in host-pathogen interaction experiments to quantify the killing capacity of macrophages. Macrophages are key components of the innate immune system as they engage in protecting the host by effective oxidative killing of C. albicans. To determine whether GCN5 is required for intracellular survival of fungal cells after phagocytosis, we infected BMDMs with wild type, gcn5∆/∆ and restored strains using an MOI of 10:1 (Candida to BMDMs). Surviving fungal cells were quantified by CFU-counting after 2 h and 4 h of infection. Strikingly, gcn5∆/∆ were much more sensitive to killing by BMDMs when compared to wild type cells or GCN5 reconstituted strains (Fig. 6C). Thus, loss of GCN5 significantly reduced fungal viability inside BMDMs.
The impaired survival in macrophages prompted us to inspect the rate and degree of fungal phagocytosis using both fluorescence microscope and FACS-based detection of macrophage-internalized Alexafluor-488-labelled Candida cells (Fig. 6D,E). BMDMs showed enhanced engulfment and phagocytosis of fungal cells lacking GCN5 when compared to wild type cells (Fig. 6D). These data are also consistent with the result showing abolished induction of SOD5 in cells lacking Gcn5 (Fig. 4A), explaining the increased oxidative killing of fungal cells by www.nature.com/scientificreports www.nature.com/scientificreports/ BMDMs (Fig. 6E). Taken together, this works establishes that the fungal Gcn5 histone acetyltransferase is essential for fungal virulence and that pathogenicity is regulated through several independent mechanisms, including adhesion, antifungal susceptibility and cell wall architecture and a strikingly increased susceptibility to killing by host-derived oxidative stress. The data establish Gcn5 as a new potential target gene for antifungal drug discovery.

Discussion
The prevalence and mortality of invasive fungal infections has been concerning. Importantly, the dramatic increases for fungal pathogens such as Candida glabrata in the past years, and the staggering increase in multidrug-resistant Candida auris infections, make fungal infections a global healthcare threat 1,79 . Therapeutic treatment options are limited to a small armory of antifungal drugs, some of which can lead to antifungal resistance or show restricted species-specificity. Hence, there is need to develop new and improved antifungal drugs, including an expansion of available chemical entities that could target new and understudied genes 80 . Compelling evidence indicates that enzymes that control chromatin modification could constitute a novel family of antifungal target genes, since they participate in numerous pathophysiological processes that regulate virulence traits 40 . Our work and data reported earlier 56 , are entirely consistent, and demonstrate that C. albicans Gcn5 is required for the hyphal formation and virulence in murine model of candidiasis. Here, we show that C. albicans Gcn5 controls morphogenesis and susceptibility to killing by phagocytes such as macrophages. Moreover, loss of Gcn5 is accompanied by Efg1-controled susceptibility to fungicidal caspofungin, as well as differentially altered MAPK signaling, which controls cell wall remodeling, adhesion and filamentation.
The Gcn5 histone lysyl acetyltransferaseis an intrinsic component of evolutionary conserved transcriptional co-activator complexes, such as SAGA, ADA and SLIK 46 , which regulate several fundamental processes [81][82][83] in yeast [84][85][86] and humans 45,46 . The knowledge about fungal Gcn5 primarily comes from studies in baker's yeast 50,81,87 and other fungal species 88,89 . Here, we use global transcriptomics and phenotypic profiling to identify regulatory networks and signaling pathways whose control engages Gcn5 in the human pathogen C. albicans. Although several phenotypes emerging upon deletion of GCN5 in C. albicans phenocopy those seen in yeast [84][85][86] , our data suggest that Candida spp Gcn5 may regulate additional networks required for virulence.
Gcn5-mediated regulation affects signaling pathways that are essential for controling morphogenetic changes such as filamentation as well as virulence. Hypersensitivities to caspofungin and other stress agents most likely arise as a consequence of altered cell wall composition and dysregulated glucan synthesis enzymes such as Fks1 and Fks2, Fks3. Interestingly, while we observed a strong repressive impact on FKS1 in gcn5∆/∆ cells, compensatory adaptive mechanisms may drive FKS2 and FKS3 upregulation. However, we do not know whether this regulation is a direct or indirect mechanism activated upon loss of Gcn5. While C. glabrata FKS1 and FKS2 are functionally redundant, a recent report suggests that C. albicans FKS2 and FKS3 negatively regulate FKS1, which modulates echinocandin susceptibility 90,91 . These data suggest severe perturbations of cell wall homeostasis derived from loss of Gcn5 control cell wall architecture and function through dysregulated MAPK signaling 92 . The Hog1 and Cek1 pathways engage in molecular cross-talk 24 and lack of Gcn5 indeed affects both pathways C. albicans.
Moreover, Gcn5 affects the synthesis of cell wall glucan and chitin through the Mkc1 cell integrity pathway 27 . Hence, our data suggest that Gcn5 differentially regulates MAPK signaling events, thereby altering cell surface composition. Indeed, the cell wall superoxide dismutase Sod5 is dramatically downregulated upon Gcn5 loss. Moreover, Gcn5 loss has a severe impact on adhesion as indicated by the dysregulation of the major Als1 adhesin, and the inability to induce Fks1 upon caspofungin stress. This dual regulation perhaps creates synthetic lethality under stress conditions, which could explain the attenuated virulence and dramatically impaired fitness in the host.
The prevailing notion that KATs in general, and Gcn5 specifically, functions as classical co-activators must be viewed with caution, since our RNA-seq data demonstrate that Gcn5, and possibly other Candida KATs, might also act as co-repressors, given the large number of about 235 genes upregulated after loss of Gcn5. For example, the elevated glucan and chitin contents imply a possible direct or indirect repressor function of Gcn5 by directly or indirectly modulating a positive transcriptional regulator required for cell wall homeostasis or adhesion. Further, the activation of both Cek1 and Hog1 MAPK pathways in the absence of stress may arise from constitutive cell wall alterations triggered by ectopic signaling. We speculate that Gcn5 alters chromatin remodelling in regulatory networks through its ability to acetylate histone H3 lysines (K14, K9, K18, K23, K27 and K36), as well as histone H4 and H2B 93 when cells encounter stress or immune defense. However, it is tempting to further speculate that Gcn5 could also acetylate non-histone proteins, including signaling components of MAPK pathways or dedicated transcription factors to regulate their activity. Alternatively, Gcn5 could control the chromatin landscapes in promoters in virulence genes in cells that encounter active host immune defense 63,94 . Indeed, our initial characterization of Gcn5-associated proteins by native co-immunoprecipitation, and the identification of the Gcn5-dependent acetylome, suggests that Gcn5 interacts and acetylates several hundred target proteins, including numerous components of the ADA, SLIK, SAGA transcriptional regulatory complexes (Shivarathri et al., unpublished data). The RNA-seq data also reveal that Gcn5 modulates additional regulatory networks governing ergosterol lipid biosynthesis, oxido-reduction and cell adhesion, ABC and MFS transporter expression and antifungal resistance 95 , as well as iron homeostasis as in C. glabrata 96 . Interestingly, iron homeostasis is tightly linked to cell wall architecture, and Gcn5 regulates expression of Rbt5 that affects cell wall architecture by regulating adhesins and the entire surface proteome [70][71][72][73][74] . In addition, Gcn5 selectively controls hyphal regulator genes such as EFG1, TEC1, but also hyphal genes, including HWP1 and ECE1. Of note, expression of NRG1 and TUP1 encoding hyphal repressors was not affected, implying that Gcn5 regulates hyphal elongation and filamentation primarily through Efg1-mediated control. In addition, deletion of GCN5 considerably diminishes expression of ergosterol biosynthesis genes, including ERG3, ERG250, ERG11, ERG13 and the binuclear Zn 2 -Cys 6 transcription factor UPC2 97,98 involved in the sterol uptake in yeast and Candida 97 . Hence, Upc2 may sense altered www.nature.com/scientificreports www.nature.com/scientificreports/ membrane lipid permeability, which is also linked to the overexpression of membrane transporters [99][100][101][102] . Indeed, gcn5 mutants show changes in non-protein-mediated membrane lipid permeability and upregulation of the Cdr1 efflux pump 103 . This may explain, at least in part, UPC2-mediated alterations in antifungal drug susceptibilities emerging upon GCN5 deletion 97 .
Moreover, ablation of Gcn5, strongly attenuates adhesion 104 , which requires Efg1 and determines immune cell interactions 16,105 . This is in line with previous reports showing that Efg1 modulates CSP tolerance through activating the major GPI-anchored cell surface adhesion protein Als1 76 . The Gcn5-mediated integration of filamentation signals also engages downstream transcription factors such as Efg1 and Tec1 29,106-108 as both EFG1 and TEC1 expression is impaired in gcn5 cells, thereby dysregulating Als1-mediated adhesion in gcn5∆/∆ cells facing CSP challenge. Of note, both ALS1 and ALS3 are also critically involved in fungal biofilm formation 109 , but a role for Gcn5 in biofilm control has not been tested.
Importantly, lack of Gcn5 significantly increases the susceptibility to killing by macrophages, because gcn5∆/∆ cells suffer from reduced fitness inside macrophages. The increased phagocytosis and or killing enables macrophages increased clearance of gcn5∆/∆ cells. This can explain for the most part the striking avirulence phenotype observed in a mouse model of systemic candidiasis, but additional unexplored mechanisms may also affect fitness of gcn5∆/∆ mutants. Of note, killing of cells lacking Gcn5 by dendritic cells is not affected, which may relate to their function as key antigen-presenting cells initiating adaptive immune responses 110 . The enhanced phagocytosis for gcn5∆/∆ may be due to increased exposure of cell surface 1,3-β-glucan, which triggers elevated ROS production by macrophages 111 . Similarly, yeast Gcn5 also modulates ROS-mediated cell death and oxidative stress 48 , although our results indicate that Gcn5-mediated ROS regulation in Candida excludes ROS produced upon caspofungin challenge 76 . Taken together, the impaired ROS scavenging by Sod5, the altered cell wall composition and adherence contribute to abrogating fungal virulence. However, the filamentation defects of gcn5∆/∆ cells might also cause reduced Sod5 induction, leading to increased susceptibility to killing. In fact, the reduced fitness may actually be a combinatorial effect, arising from defective Gcn5-mediated SOD5 regulation (the major ROS-detoxifying SOD in Candida), impaired filamentation, along with the inability to escape from the phagosome. We speculate that Gcn5 may act in concert with Efg1 that integrates environmental and morphogenetic signals from upstream MAPK cascades, as well as protein kinase A (PKA) pathways 31,75,112 .
The C. albicans genome contains at least ten catalytic subunits 40,113 with KAT activity, some of which may share redundant functions with Gcn5. For instance, other Candida KATs implicated in growth, morphogenesis and virulence include Ngg1, Esa1, as well as Hat1, Sas2 and Rtt109 36,38,39,[114][115][116] . Of note, GCN5, NGG1 and ESA1 are required for filamentation and yeast-to-hyphal growth, whereas genetic ablation of HAT1 and SAS2 induces hyperfilamentation 37,115 suggesting both positive and negative regulatory input by Candida KATs. Of note, the Gcn5-related acetyltransferase Ngs1 is required for sensing of the key filament inducer N-acetylglucosamine GlcNac. Ngs1 binds GlcNAc to activate its C-terminal acetyl transferase domain, driving promoter histone acetylation and transcription, including its promoter recruitment by the transcriptional regulator Rep1 117 .
KATs always act in concert with histone deacetylase in the regulation of both histone and non-histone targets, and like KATs, several C albicans KDACs such as Set3C, Rpd3, Rpd31, Hda1, Hos2, and Hst3 play essential roles in fungal virulence 34,40,118 . Therefore, KATs/KDACs are emerging drug targets 18,39,40,43,119 amenable to antifungal drug discovery, as the current antifungal arsenal may become insufficient owing to intrinsic and pronounced drug resistance in Candida glabrata, and broad-spectrum antifungal resistance 14,120 in emerging pathogens such as C. auris 12,121,122 . For example, the fungal KDAC inhibitor MGCD290, proved active in combination with fluconazole and echinocandins against drug-resistant Candida, as well as filamentous fungi 123,124 . Importantly, several Candida KATs target specific lysine residues on histones tails which are either absent or not modified by mammalian KAT orthologues. This suggests that fungal KAT inhibitors are unlikely to adversely affect mammalian KATs, making them especially suitable for drug discovery, and possibly useful for combination therapies 40,125 .
Taken together, our data suggest that Gcn5 holds new promises for therapeutic options to treat of invasive fungal infections, as Gcn5 controls several independent cellular and molecular pathways, all of which are critical for fungal fitness under immune surveillance. Interestingly, emerging evidence indicates that non-histone targets of KATs and perhaps KDACs could also play pivotal roles in fungal virulence and drug resistance 40,119 , opening yet another new window of opportunity for exploiting KATs/KDACS and their protein targets for antifungal drug discovery. strains and growth conditions. Candida albicans wild type and mutant strains were grown either in rich medium (YPD; 1% yeast extract, 2% peptone, and 2% dextrose) or in synthetically defined yeast nitrogen base (SC; 0.67% yeast nitrogen base and 2% dextrose) medium at 30 °C with shaking at 200 rpm. Logarithmic phase cells were obtained by growing overnight cultures in fresh yeast YPD/nitrogen base medium for 4 h at 30 °C. 200 µg/ml Nourseothricin was used as a selection marker for C. albicans. Bacterial strains were grown at 37 °C in LB medium containing either 120 µg g/ml ampicillin or 40 µg g/ml chloramphenicol. 2% agar was added to the plates. For bacterial transformations, SOC (add 20 mM glucose, 10 mM MgSO 4 , 10 mM KCl, 2.5 mM KH 2 PO 4 to LB medium) media was used. The strains, plasmids, and primers used are listed in Supplementary Tables S1-S3. www.nature.com/scientificreports www.nature.com/scientificreports/ were amplified using appropriate primers (Supplementary Table S1) covering approximately 50 bp of the homologous region to NAT1 marker cassette (to replace the whole coding region of GCN5) and YEP352 plasmid which is containing ampicillin resistance marker and E. coli origin of replication. The FRT-FLP-NAT1-FRT cassette was amplified by using the plasmid pSFS3b. PCR amplified upstream, downstream fragments, NAT1 marker cassette and YEP352 plasmids were cloned in vivo recombination either in E. coli EL350 cells or DH5α resulting in gcn5∆ plasmid. Purified restriction digested gnc5∆ plasmid was used to transform into C. albicans wild type cells. Two rounds of integration and excision generate homozygous gcn5∆/∆ mutant. Transformation of C. albicans was done via electroporation exactly as described previously 57 . Growth and phenotypic characterization. For monitoring growth curves, overnight-grown Candida cultures were inoculated into YPD with or without caspofungin or into SC medium containing various carbon sources as indicated (initial OD 600 of 0.1). Absorbance was recorded after various intervals and OD 600 values were plotted versus time. Phenotypic characterization of Candida albicans mutants was done via serial-dilution spotting analysis on agar plates. Equal volumes (3 µl) of 10-fold serial dilutions of logarithmically growing C. albicans strains were spotted onto YPD plates containing different stress agents such as temperature stress (37 °C), serum (10%), caffeine (10 mM), SDS (0.05%), Congo Red (150 µg/ml), caspofungin (CSP, 100 and 150 ng/ml), hydrogen peroxide (H 2 O 2 , 4, 5 and 6 mM), Itraconazole (ITZ, 0.02 µg/ml), fluconazole (FLC, 2, 4 and 8 µg/ml), ketoconazole (KTZ, 0.05 µg/ml) and voriconazole (VCZ, 0.02 µg/ml) along with vitamin C (VitC, 25 mM) when indicated. Colony growth was scored after 48 h and compared to the YPD control plate. SC media plates containing different carbon sources such as glucose (2%), ethanol (2%), citric acid (2%), sodium acetate (2%), and glycerol (2%) were used to assess the utilization of alternative carbon sources.

Calcoflour white staining and microscopy. To stain cell wall chitin with Calcoflour White (CFW;
Fluorescent Brightener 28, Sigma), 1 ml aliquots of logarithmically growing cells were fixed in 4% p-formaldehyde for 1-2 hours, stained with CFW 1 mg/ml for 5 min. Differential Interference Contrast (DIC) and UV light images (UV) of the same cells are shown at 60x magnification. Scale bar = 5 µM. To inspect hyphal morphologies, 1 ml of logarithmically growing cells at 30 °C and 37 °C in YPD supplemented with 10% FCS were washed twice with PBS and fixed in 4% p-formaldehyde for 2 hours. Fixed cells were washed and images were taken with an LSM 700 Zeiss Confocal microscope at 60x magnification. Scale bar = 5 µM.
Ros assay. Intracellular reactive oxygen species (ROS) was measured as described previously 37 using the following modifications. Overnight grown Candida cultures were inoculated in SC medium containing 10% glucose at an initial OD 600 of 0.3 and cultured at 30 °C for 1 h. The cultures contained 10% instead of 2% glucose to avoid CSP-induced flocculation 76 . Further, cultures were incubated with 20 mM dihydroethidium (DHE, Invitrogen) for 1 h, followed by caspofungin treatment (CSP, 150 ng/ml) for 150 min. Then, cells were washed once with water and samples were analysed by flow cytometry using FACS Calibur (BD Biosciences) at FL3-H channel. A minimum of 10,000 events were recorded for each sample. The data were analysed using Flowjo software (Flowjo LLC) and expressed mean relative fluorescence units from three independent different experiments.
Flourescein diacetate (FDA) uptake assay. Candida strains were grown in the presence or absence of caspofungin (CSP, 50 ng/ml) for 4 h. About 5 × 10 6 cells were resuspended and washed twice in 1 ml of FDA buffer before supplementing with 50 nm FDA. A 200 µl volume of cell mixture with or without FDA was added to an optical-bottom 96-well plate. The kinetics of FDA uptake was recorded every 5 min for 30 reads or until saturation was reached with simultaneous shaking of samples on the H1 Synergy plate reader with an excitation and emission wavelengths 485 and 535 nm, respectively. Data represent the mean fluorescence intensity over time. The slope rate was calculated using GraphPad Prism software.

Western blot analysis.
Logarithmically growing Candida cultures were washed once with ice-cold water and whole-cell extracts were prepared by trichloroacetic acid (TCA) method as described previously 126 . Extracts corresponding to 1 OD 600 (1 × 10 7 cells) were fractionated by 12% SDS-Page and blotted for proteins as indicated. Signals from the same whole cell extracts were detected using antibodies for total and active phosphorylated MAP kinases. The commercial antibodies recognized Mkc1 and Cek1 (p44/42 MAPK Erk1/2, Cell Signaling), and Hog1 (y-215, Santa Cruz), and phosphorylated Mkc1-P and Cek1-P (Phospho-p44/42 MAPK (Erk1/2), Cell Signaling) and Hog1-P (Phospho-p38, Cell Signaling). Reprobing with PSTAIR antibody (Sigma) recognizing Cdc28 served as a loading control. Protein bands on the nitrocellulose membrane were visualized using an Odysee ® CLx scanner (Li-Cor ® ). Quantification of the protein band intensity was performed by using image studio software (LI-Cor ® ).

Quantification of cell wall components by flow cytometry. Quantification of cell wall components
by flow cytometry was performed as described previously 60 . Briefly, logarithmically growing cells of SC5314 wild type (wt), homozygous deletion (gcn5∆/∆) and restored (gcn5∆/∆::GCN5) strains were washed and stained with Concanavalin A-conjugated Texas Red, Dectin-1/Fc + 488 and CFW to decorate mannans, glucan and chitin, respectively. These triple-stained cells were measured in a BD Fortessa flow cytometer (BD biosciences) to quantify the amount of chitin, glucan and mannan using the BV421 (violet 405 nm, 50 mW power), FITC (blue 488 nm wavelength, 50 mW power) and Texas Red (red 640 nm wavelength, 40 mW power) lasers, respectively. (2019) 9:9445 | https://doi.org/10.1038/s41598-019-45817-5 www.nature.com/scientificreports www.nature.com/scientificreports/ A minimum of 10,000 events were recorded for each sample and the data were analysed using Flowjo software (Flowjo LLC). Unstained and single-stained samples served as controls and the data expressed as the mean fluorescence intensity from three independent experiments.

Quantitative Real-time PCR (qPCR) analysis, RNA-seq analysis and bioinformatics. Total
RNA isolation, cDNA synthesis and qPCR analysis was done as described previously 37 . The efficiency-corrected ΔΔCt method was used to quantify mRNA expression levels of a target gene transcript in comparison to a reference gene transcript 127 . The mRNA of the gene associated with Topoisomerase II (PAT1) was used as a reference gene 128 . GraphPad Prism software was used to perform statistical analyses of independent biological replicates as indicated.
For RNA-seq analysis, 5 μg DNase-treated total RNA was subjected to mRNA purification using Dynabeads mRNA purification kit (Invitrogen). Remaining rRNA contamination was checked in the bioanalyzer using Agilent RNA 6000 Pico kit (Agilent). Purified mRNA was subjected to fragmentation using the NEBNext ® Magnesium RNA Fragmentation Module (New England Biolabs) and purified with RNeasy Plus Mini kit (Qiagen). Then, first-strand cDNA was synthesized with 50 ng/μl random hexamer primers using SuperScript ® III First-Strand Synthesis System for RT-PCR (Invitrogen). dNTPs were eliminated by purifying with MiniQuickSpin DNA Columns 129 and subjected to second strand cDNA synthesis followed by clean-up with Minielute Reaction cleanup kit (Qiagen). cDNA concentration was measured using Quant-iT Picogreen dsDNA Reagents (Invitrogen) in a NanoDrop Fluorospectrophotometer ND-3300 (Thermo Fisher). cDNA samples were further processed for library preparation and sequenced on a HiSeq. 2500 (Illumina) at the Next Generation Sequencing Core Facility, Vienna Biocenter Core Facilities (VBCF, https://www.vbcf.ac.at/facilities/ next-generation-sequencing). Three biological replicates for each time-points of wild type (wt) treated with caspofungin (CSP, 0, 15 and 45 min), as well as homozygous deletion (gcn5∆/∆) and two biological replicates for the remaining two time-points of gcn5∆/∆ (CSP, 15 min and 45 min) were sequenced. The RNA-seq data has been deposited in Gene Expression Omnibus (GEO) under the accession number GSE123412.
The bioinformatics analysis pipeline relied on the C. albicans genome Assembly 22 (http://www.candidagenome.org) to map sequence reads using TopHat, allowing for uniquely mapped reads 130 . HTSeq 131 with union mode was used to assess the read counts using a reference annotation (C_albicans_SC5314_version_ A22-s07-m01-r70; http://www.candidagenome.org). To identify the differentially expressed genes, read counts utilized DESeq2 R package 132 with an adjusted P-value cut-off ≤ 0.05.
Heat map of hierarchical clustered and differentially expressed genes was generated using data mining tool Orange3 133 . Hierarchical clustering used the Euclidian distance and average cluster linking. The resulting dendrograms show the degree of similarity in gene expression. Venn diagrams were generated using Venny 2.1 (http:// bioinfogp.cnb.csic.es/tools/venny) 134 . Gene ontology (GO) annotations were used the GO slim mapper tool (http://www.candidagenome.org/cgi-bin/GO/goTermMapper), and the online bioinformatics tool Fungifun2 (https://elbe.hki-jena.de/fungifun/) 135 . Adherence assays. Adherence on polystyrene-coated plates was measured by crystal violet staining as described previously 136 with minor modifications. Samples containing 2 × 10 7 Candida cells in YPD were loaded into flat-bottomed 96-well microtiter plate (Corning) and incubated for 4 hours at 30 °C. The culture medium was aspirated and cells washed once with PBS to remove non-adherent Candida. Plate wells were allowed to dry at room temperature after 15 min methanol fixation and cells were stained for 5 min with 200 µl of 1% crystal violet (v/v). Then, cells were washed gently once with water, followed by the addition of 200 µl of 33% acetic acid for destaining, before absorbance was measured at 570 nm. Data are expressed as mean of absorbance of Candida strains from three independent technical as well as biological replicates. phagocytosis assay. Flow cytometry-based phagocytosis assays were performed as described previously 36 .
Wild type (wt), homozygous deletion (gcn5∆/∆) and restored (gcn5∆/∆::GCN5) cells were grown overnight to an OD 600 of 1 and washed twice with PBS. Cells were stained with 10 mg/ml Alexa Fluor 488 (Life Technologies) in 100 mM HEPES buffer (pH 7.5) for 60 min at 30 °C with shaking in the dark. Stained Candida cells were washed thrice and resuspended in HEPES buffer, before infecting primary bone marrow-derived macrophages (BMDMs) at a multiplicity of infection (MOI) of 2:1. Phagocytosis was allowed and for 45 min at 37 °C and 5% CO 2 , before stopped by chilling samples on ice. Cells were fixed in 1% p-formaldehyde and stained with 0.4% trypan blue to quench the fluorescence of extracellular or adherent Candida. Control samples were kept on ice throughout the experiment. Intracellular phagocytosed Candida cells were quantified by flow cytometry analysis with FL1-H on a FACSCalibur (BD Biosciences). The data were analysed using Flowjo software (Flowjo LLC) and expressed as mean percentage of phagocytosis from three biological replicates.
Fluorescence microscopy-based phagocytosis assays were performed as described previously 137 , using minor modifications. BMDMs were infected with Alexafluor 488-labeled Candida strains at an MOI of 1:5 for 45 min, washed with PBS and further stained with 100 nM LysoTracker Red (DND-99) for 30 min prior to fixation. After 20 min in 3.7% formaldehyde fixation medium, cell were permeabilized for 15 min with 0.7% Triton X-100. DNA in permeabilized cells was stained with DAPI (49,6-diamidino-2-phenylindole) for 30 min. Microscopic slides were examined with a laser scanning microscope (LSM 700; Carl ZEISS). Images shown are representative overlays of all four channels red, blue, grey and green to indicate staining of lysosomes, nucleus, Differential Interference Contrast (DIC) and Candida albicans, respectively. Scale bar = 5 µM. www.nature.com/scientificreports www.nature.com/scientificreports/ (mDCs) were isolated, and cultivated exactly as described before 138 . Survival of C. albicans in BMDMs and mDCs was quantified as described previously 63,139 using an MOI of 10:1 (fungi to macrophages). Fungal cells were harvested 2 h and 4 h post infection and viability quantified by cfu-counting of samples on YPD. Survival was calculated as percentage of viable cfus after 48 h infection by comparing with uninfected Candida strains.

Immune cells and mouse strains.
For virulence experiments, mouse infections were carried out as described previously 36,78 . Briefly, C. albicans strains were grown from frozen stocks overnight to an OD 600 of around 1, washed twice and finally resuspended in PBS. For infection, 1 × 10 5 Candida cells per 21 g body weight were injected into mice via the lateral tail vein. For survival experiments, mice were monitored for 15 days, including the recording of weight loss. A group of five mice were used for each strain. Statistical analysis was carried out using the GraphPad Prism software (Graphpad Software Inc.). Mouse survival curves used the log-rank (Mantle-Cox) test. For statistical analysis unpaired two-tailed Student's t-test with 95% confidence intervals were used. P-values such as *p-value < 0.05; **p-value < 0.01; ***p-value < 0.005 were considered significant.

Data Availability
The raw datasets and annotated gene expression analysis files have been deposited in the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) database under the accession number GSE123412.