The proteasome regulator Rpn4 controls antifungal drug tolerance by coupling protein homeostasis with metabolic responses to drug stress

Fungal pathogens overcome antifungal drug therapy by classic resistance mechanisms, such as increased efflux or changes to the drug target. However, even when a fungal strain is susceptible, trailing or persistent microbial growth in the presence of an antifungal drug can contribute to therapeutic failure. This trailing growth is caused by adaptive physiological changes that enable the growth of a subpopulation of fungal cells in high drug concentrations, in what is described as drug tolerance. Mechanistically, antifungal drug tolerance is incompletely understood. Here we report that the transcriptional activator Rpn4 is important for drug tolerance in the human fungal pathogen Candida albicans. Deletion of RPN4 eliminates tolerance to the commonly used antifungal drug fluconazole. We defined the mechanism and show that Rpn4 controls fluconazole tolerance via two target pathways. First, Rpn4 activates proteasome gene expression, which enables sufficient proteasome capacity to overcome fluconazole-induced proteotoxicity and the accumulation of ubiquitinated proteins targeted for degradation. Consistently, inhibition of the proteasome with MG132 eliminates fluconazole tolerance and resistance, and phenocopies the rpn4Δ/Δ mutant for loss of tolerance. Second, Rpn4 is required for wild type expression of the genes required for the synthesis of the membrane lipid ergosterol. Our data indicates that this function of Rpn4 is required for mitigating the inhibition of ergosterol biosynthesis by fluconazole. Based on our findings, we propose that Rpn4 is a central hub for fluconazole tolerance in C. albicans by coupling the regulation of protein homeostasis (proteostasis) and lipid metabolism to overcome drug-induced proteotoxicity and membrane stress.


Introduction
The yeast Candida albicans causes life-threatening infections in patients with reduced immune function [1]. Treating these infections remains challenging, as there are only a handful of clinically useful antifungal drugs [2]. New antifungal drugs are being developed [3], but it is also important to understand how fungi overcome current antifungal agents so that improved treatment strategies could be devised [4]. In this report, we focus on understanding how C. albicans overcomes fluconazole, an azole drug that is widely used for the treatment of fungal infections [2].
Azoles inhibit Erg11, a cytochrome P450 enzyme in the biosynthetic pathway of the fungal lipid ergosterol. Inhibition of Erg11 causes membrane stress because ergosterol is depleted and toxic sterol metabolites accumulate. The azoles drugs are active against a range of fungal pathogens, however the evolution of drug resistance can cause treatments to fail [4]. Azole resistance is relatively well understood. It involves classic drug-resistance mechanisms: activation of drug efflux, increased levels of the drug target (in this case Erg11) and mutations in the target that reduce drug binding, reviewed in [4]. C. albicans can further overcome azole-induced growth inhibition by a different set of mechanisms that have collectively been termed "drug tolerance" [4][5][6][7]. Tolerance mechanisms reflect a heterogenous behaviour of cell populations, whereby some (but not all) cells are able to overcome drug-induced growth arrest and then divide at high drug concentrations [4,5,8]. As such, tolerance describes trailing or persistent growth in the presence of an antifungal drug, which could contribute to poor patient outcomes [4,5,7,8].
Although it is likely that there is some mechanistic overlap between azole resistance and tolerance, these two responses are considered to be distinct because they can be uncoupled in response to chemical inhibitors or mutations in certain cellular pathways [4]. For example, inhibitors of stress signalling decrease fluconazole tolerance by reducing the fraction of the cell population that can grow in high drug concentrations. However, drug resistance remains the same in the presence of these inhibitors, since the minimal inhibitory drug concentration (MIC) for the fungal cell population as a whole is unchanged [4,8].
The mechanisms responsible for azole tolerance are incompletely defined, although some patterns are emerging. For instance, tolerance of C. albicans to fluconazole is related to the ability of some cells in the population to limit the intracellular drug concentration [8]. Another proposed mechanism of azole tolerance is linked to changes in membrane composition, whereby increased sphingolipid biosynthesis can compensate for reduced ergosterol biosynthesis in the presence of azoles [8][9][10]. Additionally, several stress responders, such as the chaperone Hsp90, calcineurin, PKC (protein kinase C) and TOR (target of rapamycin) have been reported to modulate azole tolerance [8,[11][12][13][14], reviewed in [4]. Identifying the full spectrum of azole tolerance mechanisms should build the knowledge base towards designing improved treatments for C. albicans infections.
In this study we reveal an important role for the proteasome and its regulator, the transcription factor Rpn4, in fluconazole tolerance by C. albicans. We show that fluconazole tolerance depends on the cell's ability to overcome proteotoxic stress and the accumulation of misfolded proteins caused by fluconazole. Rpn4-dependent transcription of proteasome genes provides sufficient proteasome capacity to ensure that protein homeostasis (i.e. proteostasis) is maintained in drug. Rpn4 further contributes to the expression of ergosterol and heme biosynthesis genes, which are needed in response to inhibition of ergosterol biosynthesis by fluconazole. Collectively, our results show that Rpn4 coordinates the cellular response to fluconazole by mitigating proteotoxicity and membrane stress. Our study identifies Rpn4 and the proteasome as targets for disabling fluconazole tolerance to improve drug efficacy.

Roles of Rpn4 in stress responses, morphogenesis and immune interactions of C. albicans
The transcriptional activator Rpn4 is best characterised in the model yeast Saccharomyces cerevisiae, where it is required for activation of the proteasome genes [15]. Its roles in C. albicans have not been studied in any detail. The C. albicans RPN4 gene is activated in the core stress response [16]. Therefore, we started by asking if the rpn4Δ/Δ mutant has any stress susceptibility phenotypes. In plate assays, rpn4Δ/Δ displayed wild type susceptibility to most of the tested stressors including cell wall, membrane, osmotic and respiratory stress (S1A Fig). The exceptions were as follows. The mutant displayed mild susceptibility to hydrogen peroxide and DMSO at 30˚C and tunicamycin at 37˚C (S1A Fig). The strongest phenotype of rpn4Δ/Δ was its susceptibility to fluconazole at 37˚C (S1A Fig). The mutant also had a moderate growth defect with a~15% slower growth rate in liquid medium (S1B-S1D Fig). The growth and stress phenotypes were generally consistent between two independent deletion clones (x and y), and were complemented when the wild type RPN4 gene was re-introduced into the mutants (S1A-S1D Fig).
C. albicans grows in distinct cellular morphologies depending on environmental conditions [17]. Cultures of rpn4Δ/Δ cells displayed a higher proportion of filamentous morphologies (hyphae and pseudohyphae) under conditions in which wild type cells were predominantly in yeast form (S2A and S2B Fig). The mutant formed normal hyphae in vitro in conditions that trigger hyphal formation (S2C Fig). The mutant also formed normal hyphae upon phagocytosis by macrophages (S2D Fig, S1 Movie). These data suggest that Rpn4 is required for maintaining yeast morphology, while it does not have a major role in hyphal growth.
Hyphal formation enables the escape of C. albicans from macrophages by promoting host cell lysis via multiple mechanisms [18][19][20]. In addition to hyphae-induced lysis, macrophages further die due to metabolic stress caused by C. albicans infection, which is triggered by rapid depletion of glucose upon fungal growth [21]. Macrophage infection rates were similar for the C. albicans-infected macrophages become dependent on glucose due to their increased glycolysis [21]. The shift of macrophages to increased glycolysis can be assessed by measuring transcriptional signatures, such as the levels of glycolytic enzymes and the glucose importer Glut1. Infection of macrophages with the rpn4Δ/Δ mutant induced these transcriptional signatures of increased macrophage glycolysis (S3D Fig). Therefore, the slower cell death of rpn4Δ/ Δ-infected macrophages is not explained by changes to immune cell metabolism. Rather, our data indicates that the slower growth and metabolic changes in the rpn4Δ/Δ mutant cause slower glucose depletion by C. albicans, which in turn prolongs macrophage viability.
Taken together, our phenotypic characterisation shows that Rpn4 contributes to drug responses, morphogenesis and immune cell interaction of C. albicans. The strongest phenotype of rpn4Δ/Δ that we detected was its hyper-susceptibility to fluconazole (S1A Fig).

Rpn4 has a prominent function in fluconazole tolerance and a more modest role in resistance
Our rpn4Δ/Δ mutant clones were obtained from the Homann transcription factor mutant library [22]. Previous screens of that library are consistent with our data in that they also reported fluconazole susceptibility for rpn4Δ/Δ [22,23]. However, detailed analyses of this phenotype have not been performed. As discussed in the Introduction, C. albicans overcomes fluconazole by a complex set of processes that include two distinct components: resistance and tolerance [4]. Therefore, to precisely decipher the function of Rpn4 in fluconazole susceptibility, we utilised disk diffusion assays which allow for the determination of drug resistance (based on the radius of the zone of inhibition or "RAD") as well as drug tolerance (based on fungal growth within the zone of inhibition measured by the fraction of growth or "FoG") [8,24]. FoG is quantified relative to the maximum growth achieved by the strain under study on that same plate [24]. As such, this accounts for any differences in growth rates between strains when comparing FoG levels.
At 37˚C, rpn4Δ/Δ showed somewhat increased RAD (Fig 1A and 1B). However, no statistical significance was found for the comparisons of the wild type to the two rpn4Δ/Δ deletion clones, although one of the mutant clones (clone x) has a significantly different RAD to its complemented strain (Fig 1A and 1B). In contrast to the relatively minor effect on RAD, there was a significant and large reduction in FoG in both rpn4Δ/Δ deletion clones (Fig 1A and 1B). Therefore, the rpn4Δ/Δ mutant displays reduced fluconazole tolerance. Complementation with wild type RPN4 restored FoG for both rpn4Δ/Δ deletion clones (Fig 1A and 1B). We also assessed these phenotypes at 30˚C and found that the low FoG phenotype of rpn4Δ/Δ was less pronounced than at 37˚C although the trend was there, and RAD was increased indicating reduced resistance (Fig 1A and 1B). The expression of the RPN4 gene was elevated in "tolerant" cells growing within the zone of inhibition compared to those growing at the edge of the plate (Fig 1C). However, we could not detect higher Rpn4 protein levels in tolerant cells (Figs 1D and S4), possibly due to Rpn4 being a short-lived protein degraded by the proteasome [25]. We next measured fluconazole susceptibility of rpn4Δ/Δ by using the CLSI method, which quantifies growth in liquid RPMI medium at 35˚C in the presence of increasing drug concentrations. This analysis showed a two-fold lower fluconazole MIC for rpn4Δ/Δ after 24 h of growth (0.5 μg/ml compared to 1 μg/ml for the wild type and complemented strains) (Fig 1E). The MIC after 24 h of growth is a measure of drug resistance [4]. Thus, this result is consistent with a modest role for rpn4Δ/Δ in fluconazole resistance as determined by disk diffusion assays Rpn4 regulates fluconazole susceptibility, with a major role in tolerance. A. Fluconazole disk diffusion assays were performed with 25 μg fluconazole disks for wild type (WT), rpn4Δ/Δ and complemented strains. Plates were incubated at 30˚C or 37˚C for 2 days. Six independent experiments were performed and gave equivalent results. One representative experiment is shown. The top panel (without the disk) shows untreated controls (no drug). B. DiskImageR quantification of RAD 20 and FoG 20 for disk diffusion experiments shown in panel A. Data points are from six independent experiments, horizontal bars represent the mean and error bars represent the standard error of mean. * P < 0.05; ** P < 0.01; ***P < 0.001 (2-way ANOVA Bonferroni's multiple comparison test). Only significant statistical comparisons are shown. C. Top: Schematic diagram on the collection of WT 'tolerant' cells from within the zone of inhibition and 'non-tolerant' cells collected from a fluconazole disk plate grown at 37˚C for 2 days. Bottom: qRT-PCR analysis on RPN4 gene expression was shown in Fig 1A and 1B. In the CLSI method, the MIC at 48 h can be used as a measure of drug tolerance [4]. No change in growth was observed for rpn4Δ/Δ between 24 and 48 h at the lowest concentration of fluconazole used (0.25 μg/ml, which is approximately the MIC 80 for the mutant) ( Fig 1E). In contrast, the wild type and complemented strains displayed increased growth from 24 to 48 h at their MIC 80 of 0.5 μg/ml fluconazole ( Fig 1E). These data are indicative of tolerant growth occurring in wild type cells, which is absent in rpn4Δ/Δ. As such, they support a role for Rpn4 in fluconazole tolerance.

The transcriptional program regulated by C. albicans Rpn4
To identify the transcriptional program controlled by Rpn4 in C. albicans, we used RNA-seq to profile rpn4Δ/Δ versus wild type under standard conditions (no stress) and in response to fluconazole. In parallel, we genome sequenced the two rpn4Δ/Δ clones. Genome sequencing identified that both clones contain a trisomy of chromosome 7 (S5 Fig). This is in line with common aneuploidies upon generation of mutants in C. albicans, including in the Homann collection from which our mutants were obtained [26]. The fact that both of the independent rpn4Δ/Δ clones maintained an additional chromosome 7 suggest that it might be conferring an advantage. Surprisingly, our analyses of the rpn4Δ/Δ RNA-seq data showed no evidence of This suggests that Rpn4 controls this set of genes on chromosome 6.
For the RNA-seq, we used a relatively high concentration of fluconazole (3 μg/ml) to mimic drug stress experienced by tolerant cells growing in the zone of inhibition. To reduce the potential for indirect effects caused by long-term fluconazole stress, we limited the time of treatment to 30 minutes. Neither the wild type nor rpn4Δ/Δ displayed loss of viability under these treatment conditions (S7 Fig). Three independent experiments were conducted and both independent deletion clones of rpn4Δ/Δ were profiled. The transcriptomes of the two mutant clones mostly clustered together and were similar to each other when compared to the wild type, in the presence and absence of fluconazole (S6C and S6D Fig). Therefore, the x and y clone data were considered as replicates and analysed together with an applied batch effect. The entire dataset can be viewed interactively at https://degust.erc.monash.edu/degust/ compare.html?code=7626de961da6ddac4911adffc3b6e7ca#/.
Under standard growth conditions (without stress), rpn4Δ/Δ differentially expressed 1112 genes (FDR 0.05, log2 of 0.585, i.e. 1.5 fold change in gene expression) (Fig 2A, S1 Dataset). Of these, 590 genes were down-regulated and 522 genes were up-regulated. The down-regulated genes showed an enrichment for gene ontology (GO) terms related to the proteasome, proteolysis and ubiquitin-dependent proteolysis (Fig 2B, full list in S1 Dataset). Of the 33 genes performed from samples collected as described above. Data was normalized to 18S rRNA. Data points represent three independent experiments, each analysed in two technical replicates. Horizontal bars represent the mean and error bars represent the standard error of mean. ** P < 0.01 (unpaired t-test). D. Western blot analysis to detect C-terminally tagged Rpn4 (Rpn4-TAP) was performed on samples collected as described in panel C. Actin served as a loading control. Three independent experiments were performed (BR1 to BR3). The uncropped Western blots are show in S4 Fig. E. Percentage growth for WT, rpn4Δ/Δ and complemented strains at 0.25, 0.5 and 1 μg/ml fluconazole using the CLSI method, determined relative to the DMSO control. OD 600nm was measured after 24 h and 48 h of growth in RPMI media at 35˚C. The OD measurements used to determine percentage of growth are given in S3 Dataset. Data points represent three independent experiments, each independent experiment was analysed in two technical replicates. Horizontal bars represent the mean and error bars represent the standard error of mean. ***P < 0.001 (2-way ANOVA Bonferroni's multiple comparison test). Only significant statistical comparisons are shown. https://doi.org/10.1371/journal.ppat.1011338.g001

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance encoding proteasome subunits in C. albicans, 29 were down-regulated in rpn4Δ/Δ (Fig 2B, S1 Dataset). In addition, several putative proteasome and ubiquitin-related chaperones and regulators were also down-regulated (Fig 2B, S1 Dataset). The expression of the transcription factor HAC1 (which regulates the unfolded protein response) was induced ( Fig 2B).
Further to the proteasome genes, rpn4Δ/Δ cells displayed lower transcript levels for several ergosterol biosynthesis genes and their transcriptional activator UPC2 (Fig 2B, S1 Dataset). The GO term "sterol metabolic process" was enriched with a p value of 0.04995 (S1 Dataset). The enzymes of heme biosynthesis HEM13, HEM14 and HEM15 were also down-regulated (Fig 2B, S1 Dataset) (note that heme is a co-factor for ergosterol biosynthesis) [27]. In contrast, several genes needed for iron uptake, including the iron-regulated transcription factors SEF1, HAP43 and IRO1, were up-regulated ( Fig 2B). We note however that some of these ironrelated genes were expressed at low levels (e.g. CFL11) and the only GO term related to metals that was significantly enriched in the rpn4Δ/Δ down-regulated gene set was "metal iron binding" (p = 0.01433, S1 Dataset). In addition to these pathways, GO categories related to carbon metabolism (including "glycolytic process", "ATP generation from ADP" and "pyruvate metabolic process"), autophagy, vacuolar transport, organonitrogen compound metabolic process and the ESCRT complex were also enriched in the genes down-regulated in rpn4Δ/Δ (S1 Dataset).
Functional groups up-regulated in rpn4Δ/Δ cells include GO terms related to ribosome biogenesis, rRNA and ncRNA processes (S1 Dataset). Although there was no significant enrichment in GO terms related to hyphae, several cell surface proteins and hyphal-specific genes were up-regulated (S1 Dataset). This includes the genes determined by Carlisle and Kadosh to represent the core hyphae-induced transcripts [28], such as ALS3, HWP1, ECE1, HYR1, RBT5, IHD1 and the activator UME6 (S1 Dataset). Up-regulation of hyphae-specific genes likely reflects the hyper-filamentous morphology of rpn4Δ/Δ cells (S2 Fig).
We next performed ChIPseq in order to determine the direct targets of C. albicans Rpn4. Unfortunately, these attempts failed with two different C-terminal tags on Rpn4 (TAP and HA), and in both standard growth conditions (without drug stress) and following the addition of fluconazole. A possible explanation for ChIPseq failure is a transient interaction of Rpn4 with DNA. We propose this because in S. cerevisiae Rpn4 has a short half-life due to a regulatory feedback loop whereby it is degraded by the proteasome [25]. It is reasonable to assume that similar regulation could be occurring in C. albicans. As the next best thing, we performed a bioinformatic search for Rpn4 binding sites in the upstream regulatory regions of C. albicans genes (+1kb). For this, we used the C. albicans Rpn4 binding site determined by Gasch et al [29]: AGTGGCAAAN, GGTGGCAAYA, GRAGGCAAAA. This search identified 217 putative Rpn4 targets (S2 Dataset). Of these, 64 were differentially expressed in rpn4Δ/Δ (56 under basal conditions, and 8 after addition of FLC). The majority of these putative Rpn4 targets (46 out of 64) were down-regulated in the mutant (S2 Dataset), suggesting that C. albicans Rpn4 acts predominantly as a transcriptional activator. Of the proteasome genes, 18 had a Rpn4 binding site. Another 10 were identified when we relaxed the search to include less common binding sites according to the matrix reported in Gasch et al (RRWGGCAAHN) [29]. Of the Fold change in gene expression in all samples is expressed relative to the wild type without fluconazole (WT -FLC). Proteosome genes RPN6, RPN1, PRE8 and PRE1 were not included in the heat maps as they fall outside the cut off criteria. The cut-off used is log2 of 0.585, i.e. � 1.5 fold change; FDR < 0.05. C. Venn diagram and GO analysis of genes differentially expressed in WT +/-FLC and rpn4Δ/Δ +/-FLC (� 1.5 fold change; FDR < 0.05). The genes used to construct the Venn diagram and the GO terms for "Process" are detailed in S1 Dataset. The cut-off p-value for GO terms was <0.01. The unique genes up-regulated by FLC in rpn4Δ/Δ did not show any significant GO terms, except "transmembrane transport" which showed a p value of 0.0155. As there are many functionally overlapping GO terms, we did not list all but summarised them in the figure for simplicity. The full GO analysis and list of terms are shown in S1 Dataset. https://doi.org/10.1371/journal.ppat.1011338.g002

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance ERG and HEM genes only ERG7 and HEM15 had a Rpn4 binding site, but in both cases over 950 bp upstream of the start codon (S2 Dataset).

Rpn4 contributes to fluconazole susceptibility by controlling ergosterol gene expression
Next we turned our efforts to the fluconazole response. Fluconazole-treated wild type cells differentially expressed 36 genes and, as expected, activated the ergosterol biosynthesis pathway (Fig 2A (middle panel), 2B and S1 Dataset). No other major changes were observed, likely due to our short treatment time of 30 minutes. The transcriptional response of rpn4Δ/Δ upon fluconazole treatment was more extensive, with 325 genes differentially expressed when comparing treated versus untreated mutant (Fig 2A right panel and S1 Dataset). Of the 172 genes upregulated in rpn4Δ/Δ, 20 were shared with the wild type ( Fig 2C). Ergosterol biosynthesis genes were enriched in this shared group (Fig 2C). Genes up-regulated by fluconazole in rpn4Δ/Δ also displayed an enrichment for categories related to import, including "iron import into cell" (S1 Dataset). However iron import genes FTH1, FTR1, SIT1, FRP1 and FET3 were mostly expressed at low levels and/or modestly up-regulated in the rpn4Δ/Δ mutant (see data at https://degust.erc.monash.edu/degust/compare.html?code= 7626de961da6ddac4911adffc3b6e7ca#/). There were only 3 genes whose expression was reduced by fluconazole in the wild type, while in rpn4Δ/Δ there were 153 genes ( Fig 2C). Of these, hyphal genes were down-regulated (S1 Dataset). This is consistent with the fact that fluconazole represses hyphal morphogenesis [30]. Several other GO terms also showed reduced expression in fluconazole-treated rpn4Δ/Δ cells, including carbohydrate transport, biofilm formation and aggregation ( Fig 2C). There were no changes to the expression of heme biosynthesis genes in response to fluconazole, in either the wild type or rpn4Δ/Δ strains ( Fig 2B, S1 Dataset). Similarly, the proteasome genes remained expressed at low levels in rpn4Δ/Δ, in the presence or absence of fluconazole ( Fig 2B).
We next studied the ERG genes more closely, since some of them were down-regulated in rpn4Δ/Δ under basal levels but up-regulated by fluconazole (Fig 3A). This analysis showed that rpn4Δ/Δ could upregulate the ERG genes in response to fluconazole but their up-regulation in many cases was smaller than in the wild type (exceptions to this were ERG25, ERG28 and UPC2) (Fig 3B). In other words, regardless of their upregulation, many of the ERG genes and their transcription factor UPC2 remained expressed at lower levels in fluconazole-treated rpn4Δ/Δ versus fluconazole-treated wild type ( Fig 3C). qRT-PCR experiments confirmed activation of ERG27 in the wild type by fluconazole and reduced activation in rpn4Δ/Δ (Fig 3D). The complemented strains restored ERG27 transcriptional activation to wild type levels ( Fig 3D).
The proteasome is a key transcriptional target of C. albicans Rpn4 (Fig 2B). Given the lack of Rpn4 binding sites upstream of the ERG genes in our bioinformatic analysis (S2 Dataset), we were wondering if the regulation of ERG gene expression by Rpn4 is controlled indirectly, via its roles in proteasome function. To test this, we treated wild type cells with the proteasome inhibitor MG132, and then profiled ERG gene expression with and without fluconazole. As a control for MG132 activity, we show that MG132 caused an upregulation of proteasome genes RPT2 and PUP1, likely as a compensatory response to proteasome inhibition (Fig 3E). At basal levels (without fluconazole), the expression of ERG11, ERG251 and ERG27 did no change in response to proteasome inhibition with MG132 ( Fig 3E, -FLC samples). This is in contrast to the down-regulation of ERG251 and ERG11 in rpn4Δ/Δ (Fig 3A). In response to fluconazole, MG132-treated cells were able to upregulate the ERG genes ( Fig 3E). We noticed that there

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance some variability in ERG gene activation in MG132 and FLC-treated cells likely due to heterogenous stress responses. We therefore performed four independent experiments, with 3 independent cultures per group in each (i.e. 12 biological repeats). While in some experiments, MG132-treated cells showed reduced ERG gene activation in response to fluconazole (see S3 Dataset), when taken together there were no major differences ( Fig 3E). These data suggest that Rpn4 has proteasome-independent roles in ERG gene expression.
To address the functional relevance of lower ERG gene expression in rpn4Δ/Δ we made use of the fact that C. albicans is capable of sterol uptake from the medium [31] and asked if supplementing ergosterol could restore fluconazole tolerance in the mutant. Indeed, this was the case: ergosterol restored growth of rpn4Δ/Δ in the zone of inhibition and FoG 20 levels of rpn4Δ/Δ in the presence of ergosterol were the same as for the wild type (Fig 4A and 4B). Collectively, our findings indicate that one of the mechanisms by which Rpn4 regulates fluconazole tolerance is by enabling ergosterol supply via ERG gene expression.

Functions of the proteasome in fluconazole tolerance and resistance
As shown by our RNAseq data, the proteasome is a key transcriptional target of C. albicans Rpn4 (Fig 2B). Proteasome genes are down-regulated in rpn4Δ/Δ cells (Fig 2B), and there is an accumulation of ubiquitinated proteins, consistent with lower proteasome activity in the mutant (Fig 5A). The levels of ubiquitinated proteins in rpn4Δ/Δ cells were higher than in wild type cells, but lower than in wild type cells treated with the proteasome inhibitor MG132 (Fig 5A). Applying MG132 to rpn4Δ/Δ cells resulted in a further accumulation of ubiquitinated proteins ( Fig 5A). Collectively, these data show that proteasome activity is diminished in the C. albicans rpn4Δ/Δ mutant, but it is not completely inactivated.
Next we asked if fluconazole causes proteasome stress by perturbing protein homeostasis. We did this by studying the accumulation of ubiquitinated proteins in response to fluconazole. In wild type cells, there was a heterogeneous response. In two out of four biological repeats fluconazole induced a clear accumulation of ubiquitinated proteins (Fig 5B, Experiment B and C). In the other two repeats (Experiment A and D), we did not observe a convincing accumulation (Fig 5B), suggesting that cells adapted and dealt effectively with the missfolded proteins. In contrast to the wild type, in rpn4Δ/Δ cells there was a reproducible accumulation of ubiquitinated proteins in response to fluconazole above what is seen in the mutant without stress (Fig 5B). Consistent with an important role for the proteasome in fluconazole responses, MG132 impaired both resistance and tolerance to fluconazole, as shown by increased RAD and reduced FoG in fluconazole disk diffusion assays (Fig 5C and 5D). The reduction in FoG by MG132 phenocopied rpn4Δ/Δ (Fig 5C and 5D). Interestingly, unlike rpn4Δ/Δ, the ubiquitin mutant ubi4Δ/Δ showed no changes to fluconazole FoG at 37˚C, although RAD was increased (S8A and S8B Fig). The lack of fluconazole FoG phenotype of ubi4Δ/Δ is possibly due to shown in Fig 2B but here shown as fold change in gene expression relative to wild type +FLC. Cut-offs were 1.5 fold change; FDR < 0.05. D. qPCR analysis of ERG27 gene expression in WT, rpn4Δ/Δ and complemented strains. Gene expression was normalized to RDN25. Data points represent three independent experiments, each analysed in two technical replicates. Horizontal bars represent the mean and error bars represent the standard error of mean. ** P < 0.01; *** P < 0.001 (2-way ANOVA Bonferroni's multiple comparison test). Only significant statistical comparisons are shown. E. qPCRs of proteasome and ERG gene expression in response to proteasome inhibition with MG132. Wild type C. albicans was treated with 50 μM MG132 for 3 h, followed by 30 min of fluconazole (FLC) treatment (3 μg/ml). Gene expression was normalized to RDN25 and expressed relative to the average expression in the untreated wild type (See S3 Dataset). Data points represent biological repeats from 4 independent experiments. In each experiment, 3 independent colonies were cultured for each condition, leading to 12 repeats. The horizontal bars represent the mean and error bars represent the standard error of mean. ** P < 0.01; *** P < 0.0001 (2-way ANOVA Bonferroni's multiple comparison test). Only significant statistical comparisons are shown. https://doi.org/10.1371/journal.ppat.1011338.g003

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance compensation from UBI3 [32]. Unfortunately, we could not test this hypothesis as UBI3 is an essentially gene and therefore a double ubi4Δ/Δ ubi3Δ/Δ mutant could not be constructed. Collectively, the results presented in Fig 5 indicate that Rpn4-dependent regulation of the proteasome is critical for providing enough capacity for cells to deal with accumulation of misfolded proteins and proteotoxicity caused by fluconazole. The ability to maintain protein homeostasis is essential for both tolerance and resistance to fluconazole.

Discussion
Here we identify the transcriptional activator Rpn4 as a regulator of fluconazole tolerance in C. albicans. Our analysis of a range of rpn4Δ/Δ mutant phenotypes showed that the strongest defect was its hyper-susceptibility to fluconazole. Since C. albicans overcomes fluconazole by a complex set of mechanisms involving both drug resistance and drug tolerance [4], we utilised assays that distinguish between these two facets of drug susceptibility. Using these assays, we found that Rpn4 is essential for fluconazole tolerance. This was especially true at the clinically relevant temperature of 37˚C, at which tolerant growth was practically non-existent in the mutant. In contrast, Rpn4 plays a more minor role in fluconazole resistance, with only a twofold change in the fluconazole MIC for rpn4Δ/Δ. We acknowledge that rpn4Δ/Δ grows more slowly than the wild type even without stress (15% reduced growth rate) (S1B and S1C Fig). However, we argue that a generalised growth defect does not explain the large drop in fluconazole tolerance observed for rpn4Δ/Δ. First, the reduction in tolerant growth of the mutant is dramatic, while the growth defect is minor. Second, tolerant growth is measured as a fraction of growth relative to maximal growth achieved on the same plate, thereby accounting for differences in growth rates between fungal strains [24]. Instead, we propose that Rpn4-activated pathways mitigate fluconazole stress in two ways: Rpn4 activates proteasome gene expression

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance for overcoming drug-induced proteotoxicity, and it further contributes to the expression of ergosterol genes to mitigate ergosterol biosynthesis inhibition and membrane stress caused by fluconazole.

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance With respect to ergosterol biosynthesis, our transcriptome data showed that Rpn4 promotes constitutive transcription of ergosterol and heme biosynthesis genes, and is also required for full induction of several ergosterol genes by fluconazole. A role for Rpn4 in alleviating fluconazole-induced ergosterol depletion is shown by the fact that supplementation of ergosterol to rpn4Δ/Δ restores fluconazole tolerance (Fig 4). Our data is supported by a recent study showing that Rpn4 activates ergosterol and heme gene expression in another yeast pathogen, Candida glabrata [33]. However, the regulatory mechanisms are somewhat different. In contrast to what we show in C. albicans, C. glabrata Rpn4 is not required for constitutive expression of the ergosterol or heme genes, only for their induction by fluconazole [33]. Also, C. glabrata Rpn4 binds directly to the ERG11 promoter, which contains an Rpn4-recognition motif [33], but we found no Rpn4 binding sites in ERG11 or other ERG genes suggesting that in C. albicans their regulation by Rpn4 is indirect. We considered that lower proteasome activity might be the reason for reduced ERG gene expression in rpn4Δ/Δ cells. Our data argues against that possibility, as expression we detected normal of ERG transcript levels upon proteasome inhibition with MG132 ( Fig 3D). The expression of UPC2, encoding the central transcriptional activator of ergosterol biosynthesis in C. albicans, was down-regulated in rpn4Δ/Δ cells (Fig 2B). This suggests that Rpn4 might be regulating the ERG genes by activating UPC2. However, we found no Rpn4 binding sites in the UPC2 promoter. It remains to be determined how C. albicans Rpn4 regulates the expression of ERG and HEM genes.
Regulation of proteasome gene expression is the best-known function of Rpn4 in yeasts [15,25]. A role for C. albicans Rpn4 in proteasome regulation has been suggested by bioinformatic analyses, which found Rpn4 binding sites upstream of proteasome genes [29]. Our data now provides experimental evidence that C. albicans Rpn4 activates proteasome gene expression under standard growth conditions and under drug stress, and we also show that proteasome activity is reduced in rpn4Δ/Δ cells. The reduction in proteasome activity sensitizes rpn4Δ/Δ cells to fluconazole-induced proteotoxicity, as shown by an accumulation of ubiquitinated proteins in fluconazole-treated rpn4Δ/Δ cells. Moreover, the reduction in fluconazole tolerance of rpn4Δ/Δ was more pronounced at 37 than at 30˚C (Fig 1). This is also consistent with a key role of proteostasis in fluconazole tolerance, as the higher temperature of 37˚C is expected to exacerbate proteotoxicity from unfolded protein accumulation, making the loss of proteasome activity in rpn4Δ/Δ more evident. Collectively, our data demonstrates that Rpn4-dependent regulation of the proteasome is critical for providing sufficient proteasome capacity to overcome fluconazole-induced proteotoxicity. Supporting this hypothesis, proteasome inhibition by MG132 had a large effect on fluconazole tolerance as well as resistance, and phenocopied rpn4Δ/Δ for reduced tolerance (Fig 5). Rpn4 plays a role in fluconazole susceptibility in two other yeasts, C. glabrata [33] and S. cerevisiae [34,35]. As such, we propose that the roles of Rpn4-regulated proteasome functions in azole responses are conserved across yeast species.
How does Rpn4 control fluconazole tolerance via the proteasome? There are several possible scenarios. Low proteasome activity and unfolded protein accumulation in rpn4Δ/Δ could tie up the Hsp90 chaperone, preventing it from exerting its functions in fluconazole tolerance [11][12][13]36]. However, rpn4Δ/Δ cells were not hypersusceptible to the Hsp90 inhibitor geldanamycin (S1A Fig), suggesting that Rpn4 has Hsp90-independent functions. A second possibility is suggested by data from S. cerevisiae, which showed that the proteasome degrades ergosterol biosynthesis enzymes [37][38][39]. This process ensures sterol homeostasis by limiting the activity of the ergosterol biosynthesis pathway to avoid the accumulation of toxic sterol intermediates [38,39]. It is conceivable (and likely) that similar mechanisms operate in C. albicans. Therefore, we speculate that Rpn4-dependent regulation of the proteasome might control Erg enzyme levels to reduce toxic sterols, which would be particularly important in response to blockade of

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance the ergosterol biosynthesis pathway by fluconazole. Supporting this hypothesis, it has been shown that the S. cerevisiae rpn4 mutant displays higher levels of several ergosterol biosynthesis enzymes due to their reduced protein degradation [40].
How does this hypothesis fit with our data showing lower ERG gene transcripts in the C. albicans rpn4Δ/Δ mutant? We considered the possibility of feedback regulation: if the C. albicans rpn4Δ/Δ mutant displays higher Erg protein levels due to lower proteasomal degradation (similar to what has been shown in S. cerevisiae), then cells could try to compensate by reducing ERG gene transcription to restore sterol homeostasis. While we cannot exclude this possibility, our data argues against it because we show that treatment of wild type C. albicans with the proteasome inhibitor MG132 does not reduce ERG transcript levels (Fig 3). Thus, we propose that the role of Rpn4 in proteasome regulation is distinct from its role in ERG gene expression. That said, it would make sense for C. albicans to use Rpn4 to coordinate ERG gene expression with proteasome-dependent degradation of the Erg enzymes. This sort of regulation would represent an important homeostatic mechanism to ensure optimal activity of the ergosterol biosynthesis pathway, so that toxic sterol intermediates are kept in check but ergosterol is supplied to membranes to enable cell growth. Indeed, our data shows that the homeostatic mechanisms regulated by C. albicans Rpn4 are essential for overcoming growth inhibition by fluconazole, which is important for drug tolerance.
The proteasome is a therapeutic target in human diseases, and Rpn4 is a fungal-specific transcriptional activator with no direct homologs in mammals. Therefore, our results shed new light on the mechanisms of antifungal drug tolerance in C. albicans and identify possible targets for therapeutic intervention.

Ethics statement
All animal experiments were approved by the Monash University Animal Ethics Committee (approval numbers ERM14292 and ERM25488).

C. albicans strains
The strains used in this study are shown in S1 Table. Two independent deletion clones of RPN4, rpn4Δ/Δ x and y (HIS1 + LEU2 + arg4 -) were obtained from a transcriptional regulator knockout library [22], which was obtained from the Fungal Genetics Stock Centre [41]. These rpn4Δ/Δ mutant clones are homozygous deletion strains in which RPN4 was deleted by utilizing the C. dubliniensis HIS1 and C. maltose LEU2 markers [42]. The deletion clones were genotyped by PCR to confirm homozygous deletion of RPN4 and were also reversed to arginine prototrophy by insertion of C. dubliniensis ARG4 into the LEU2 locus to become HIS1 + LEU2 + ARG4 + . To construct the RPN4-complement strain, one copy of wild type RPN4 was introduced at the LEU2 locus of rpn4Δ/Δ using the C. dubliniensis ARG4 marker. In addition, a second wild type RPN4 copy was introduced at the endogenous RPN4 locus upstream of the HIS1 disruption cassette using the NAT1 selectable marker. Genome sequencing of the strains identified that both deletion clones (x and y) contained a trisomy of chromosome 7, which was however transcriptionally buffered (S5 and S6 Figs). For all of the above strains, the wild type reference SN425 (HIS1 + LEU2 + ARG4 + ), a phototrophic strain derived from SN152, was used.

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance leu2 -ARG4 + ) was constructed by introducing a TAP tag amplified from plasmids pFA-TAP-HIS1 or pFA-TAP-ARG4 by PCR to the C-terminus of RPN4 in the parental strain SN152 [43]. Both alleles were tagged. For the Rpn4-TAP strain, the wild type reference SN425 (HIS1 + LEU2 + ARG4 + ) was used. All transformants were confirmed by diagnostic PCR (for primers see S2 Table).

Growth conditions for fungal cultures
For standard growth, C. albicans strains were grown in YPD medium (1% yeast extract, 2% peptone, 2% glucose, 80 μg/ml uridine, with addition of 2% agar for plates) at 30˚C with shaking at 200 rpm. For yeast growth, overnight-cultured strains were diluted to OD 600nm of 0.2 into prewarmed YPD medium at 30˚C with shaking at 200 rpm. For hyphal growth, overnight-cultured strains were diluted to OD 600nm of 0.2 into prewarmed bone-marrow-derived mouse macrophages (BMDM) medium containing serum-free RPMI 1640 media pH 7.4 supplemented with 10 mM glucose at 37˚C with shaking at 200 rpm. Ergosterol (Sigma-Aldrich) was supplemented onto YPD plates to determine the effect of exogenous ergosterol on fluconazole susceptibility. A solvent control of 1:1 Tween80/Ethanol was included. For determining cell viability following fluconazole treatment, overnight-cultured strains were diluted to OD 600nm of 0.2 and grown to log phase (OD 600nm of 1) in YPD medium at 37˚C with shaking at 200 rpm. Cells were then treated with 3 μg/ml fluconazole or the solvent control DMSO (Sigma-Aldrich, final concentration 0.06%) for 30 min. Cultures were serially diluted and plated on YPD agar plates. Colonies were counted after 2 days of incubation at 30˚C.

Macrophage infection experiments and live cell microscopy
Bone marrow were extracted from femur and tibia bones of male or female 6 to 8 weeks old C57BL/6 mice from the Monash Animal Research Platform. Extracted monocytes were suspended in differentiation medium: RPMI 1640 pH 7.4 supplemented with supplemented with 10mM glucose, 12.5mM HEPES, 15% fetal bovine serum (Serana), 20% L-cell conditioned medium containing macrophage colony-stimulating factor and 100 U/mL of penicillin-streptomycin (Sigma-Aldrich), and differentiated into mouse bone marrow-derived macrophages (BMDMs) for 5 to 7 days, as previously described [44]. A cell scraper was used to scrape differentiated macrophages from petri dishes. Macrophages were then seeded onto 96-well microplates at a density of 5 x 10 5 cells/well and incubated overnight at 37˚C in 5% CO 2 . Before fungal challenge, macrophages were stained with 1 μM CellTracker Green CMFDA dye (Thermo Fisher Scientific) for 20 min in serum-free RPMI medium 1640 pH 7.4 lacking glucose. To prepare C. albicans for macrophage infections, single colonies were patched onto YPD plates and incubated at 30˚C overnight. C. albicans cells were scrapped off plates, resuspended in PBS and inoculated at 7.5 x 10 4 cells/ml in differentiation medium (described above). Macrophages were then infected by replacing serum-free CellTracker Green medium with the differentiation medium (as described above) containing C. albicans cells at a multiplicities of infection (MOI) of 1.5:1 for S3B Fig. Phagocytosis was allowed to proceed for 1 h, followed by washing of non-phagocytosed fungal cells with PBS. Mouse BMDM differentiation medium was then added to each well, supplemented with 10 mM glucose. To track dead macrophages, 0.6 μM DRAQ7 (Abcam) was added to the medium. Live cell imaging was performed at 37˚C in 5% CO 2 as previously described [21,44]. A minimum of 1000 macrophages were counted for each biological replicate. Imaging was performed using a Leica DMi8 with a GTC9000 camera and Leica LAS X software. Fluorescence excitation needed to image cell death nuclei (DRAQ7) was achieved using the Leica EL6000 external light source with a Y5 filter cube. All images were captured using the HC PL FLUOTAR L 20x / 0.40 Dry PH1 CORR

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance objective. The data were analysed and quantified using CellProfiler (version 2.1.1) and FIJI image analysis programme [45]. Selected microscopic positions were excluded based on counting irregularities caused by out-of-focus images. Snapshots of the live cell imaging data were obtained using FIJI.
To calculate the phagocytic index (S3A Fig), macrophages were infected with C. albicans cells at a MOI of 2:1. At 1 h post-phagocytosis, macrophages were washed with PBS to remove un-phagocytosed C. albicans and fixed with 4% paraformaldehyde. Following fixation, macrophages were permeabilised using 0.1% Triton-X 100 and then stained with 10 μg/ml calcofluor white to count the phagocytosed C. albicans cells. A minimum of 1800 macrophages were counted for each biological replicate.
For determining glucose concentrations in the macrophage infection medium (S3C Fig), infections were performed as above (MOI 1.5 of wild type or rpn4Δ/Δ C. albicans:1 macrophage) and samples taken after 1, 12 and 14 h. The concentration of glucose was determined using the glucose oxidase kit (Amplex Red, Thermofisher), following the manufacturer's instructions.

Quantitative PCR analysis of gene expression
Overnight-grown cultures of C. albicans strains were diluted to an OD 600nm of 0.2 and grown in 10 or 20 ml of YPD medium at 37˚C with shaking at 200 rpm. Total RNA was extracted using the hot phenol method as we described previously [46]. One-μg of DNase I-treated total RNA was reverse-transcribed using SuperScript III Reverse Transcriptase (Invitrogen) according to the manufacturer's instructions. qRT-PCR reactions were performed on the LightCycler 480. Data were analysed with the LinReg software (version 11.0) [47]. At least three biological repeats were performed per experiment with two technical repeats (number of repeats is indicated in the figure legends). Statistical analysis on biological repeats was performed on the normalized expression values to either the RDN25 or the 18s rRNA gene. For detecting the expression levels of RPN4 in Fig 1C, 'tolerant' wild type cells from within the zone of inhibition and 'non-tolerant' cells outside the zone of inhibition were collected from a fluconazole disk plates grown at 37˚C for 2 days and resuspended in PBS. For detecting the expression levels of ERG27 in Fig 3D, overnight-grown cultures of C. albicans strains were diluted to an OD 600nm of 0.2 and grown till log phase (OD 600nm of 1) in YPD medium at 37˚C with shaking at 200 rpm, followed by additional incubation of 30 min in 3 μg/ml fluconazole or the solvent control DMSO (final concentration 0.06%) with shaking at 200 rpm.
For testing the effects of proteasome inhibition with MG132 on gene expression, wild type C. albicans was grown over night in YPD medium. Overnight-grown cultures were diluted to an OD 600nm of 0.2 and grown in 10 ml of YPD medium to OD~1.0 (approximately 3 h) at 37˚C with shaking at 2000 rpm in the presence or absence of 50 μM MG132. Control cultures contained an equivalent amount of the DMSO solvent. After reaching the desired OD, one set of samples was centrifuged at 3000 rpm for 5 min and cell pellets frozen as no-drug controls. Another set of samples had 3 μg/ml of fluconazole added and treatment performed for 30 min, before centrifugation and freezing of cell pellets as above. RNA isolation and qPCRs were performed as described above.
For analysing the expression of macrophage glycolytic genes (S3D Fig), infections were performed as described above (MOI 3 C. albicans:1 macrophage), followed by using Triazol reagent to lyse macrophages with water after 1, 3, 6 and 9 h. Controls were uninfected macrophages. Macrophage RNA was isolates using Triazol, and reverse transcription and qPCRs performed as described above. Eef2 rRNA expression was used for normalising the data.
Sequences of all qPCR primers used are listed in S2 Table. PLOS PATHOGENS Rpn4, the proteasome and azole drug tolerance

Fluconazole disk assay and diskImageR analysis
DiskImageR is publicly available from CRAN (the 'Comprehensive R Archive Network'). Full instructions of the diskImageR pipeline are detailed in [24]. The disk diffusion plates were photographed under a consistent setting using the Phenobooth (Singer Instruments). Photograph size is standardized and the centre of the disk is located. Thirty-five millimetres radial lines are drawn every 5˚starting at the disk centre and finishing at the plate edge. This is then averaged to determine the density from the disk to the edge. Variation in plating is normalized by diskI-mageR. The plate background is established by subtracting all values from the pixel intensity of a plate with clear space behind the disk. RAD 20 values represent the distance in mm from the edge of the disk that corresponds where 20% growth reduction has occurred. FoG 20 values are calculated using the area under the curve in slices from the disk edge to the RAD 20 cut-off. This achieved growth is compared to the potential growth. Two-hundred μl of overnightgrown cultures were diluted to an OD 600nm of 0.1 (equivalent to 10 6 cells/ml) and plated onto YPD agar using 3 mm glass beads. A single 25 μg fluconazole disk (6 mm diameter, Thermo Fisher Scientific) was placed using sterile forceps onto the middle of the plate. To ensure consistent cell density plated on each plate, the same diluted cultures were also spread on YPD agar plates without a fluconazole disk. Plates were incubated at 30˚C or 37˚C for 2 days and photographed.

Microscopy
Microscopy images were taken using the EVOS FL microscope (Thermo Fisher Scientific). For S2A Fig, overnight

Western blot analysis
For detecting the levels of TAP-tagged Rpn4 protein in Fig 1D, 'tolerant' cells from within the zone of inhibition and 'non-tolerant' cells outside the zone of inhibition were collected from a fluconazole disk plates grown at 37˚C for 2 days and resuspended in PBS. The suspension was centrifuged at 14000 rpm for 3 min and washed with PBS in screw-cap tubes. Cell pellets were snap-frozen on dry ice and stored at -20˚C awaiting protein extraction. Whole-cell protein extracts were performed by adding 100 μl glass beads, 100 μl 20% trichloroacetic acid (TCA) to cell pellets on dry ice till the TCA froze. The tubes were thawed by vortexing. Glass beads were rinsed twice in 500 μl 10% TCA and supernatant was transferred to a new tube and centrifuged at 14000 rpm for 5 min. Proteins were precipitated in 1 ml ice cold acetone and resuspended in high pH Laemmli Buffer (0.0625 M Tris pH 8.8, 2% SDS, 10% glycerol, 5% mercaptoethanol, 0.01% saturated bromophenol blue), and boiled at 100˚C for 5 min. Boiled samples were loaded onto a 10% SDS-PAGE gel and transferred to a PVDF membrane at 700 mA, 300 mV using the semi-dry transfer method (Bio-Rad) for 20 min. Blocking of the membrane was performed using 5% skim milk in TBS-T (TSB with 0.1% Tween) for 1 h. For detection with the anti-TAP (Open Biosystems, Cat#CAB1991) and anti-Actin (Millipore, Cat#MAB1501) primary antibody, the membranes were incubated for 1 h at room temperature at 1:1000 dilution and 1:5000 dilution, respectively. The membranes were washed 3 times for 5 min with TBS-T followed by incubation with HRP-conjugated secondary antibodies (Sigma-Aldrich, anti-TAP

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance secondary antibody Cat#A0504 and anti-Actin secondary antibody Cat#A4416) at room temperature at 1:1000 dilution in 5% skim milk in TBS-T for 1 h. Membranes were visualized using Clarity Western ECL substrate (Bio-Rad) and exposed to medical X-ray film [48].
Testing the effects of fluconazole on the accumulation of ubiquitinated proteins: cultures of wild type and rpn4Δ/Δ C. albicans were grown overnight in YPD medium. Overnight-grown cultures were diluted to an OD 600nm of 0.2 and grown in 10 ml of YPD medium for 2 h at 37˚C with shaking at 200 rpm. After that, the no-drug samples were harvested by centrifugation at 3000 rpm for 5 min, and cell pellets snap frozen. The fluconazole samples were obtained by adding 3ug/ml fluconazole and incubating for 4 h before harvesting cells as above. Protein extracts were prepared using the TCA method as above and separated on a 10% SDS PAGE gel. Western blots were performed using the anti-Ubiquitin antibody from Santa Cruz Biotechnologies (1: 2000 dilution, 1 h incubation at room temperature), followed by the antimouse HRP secondary antibody (1: 10000 dilution at room temperature). For experiments addressing the effects of MG132 on proteasome inhibition, overnight fungal cultures were diluted to an OD 600nm of 0.2 and grown in 10 ml of YPD medium to OD~1.0 (approximately 3 h) at 37˚C with shaking at 200 rpm in the presence or absence of 50 μM MG132. Controls contained an equivalent amount of the DMSO solvent. Cells were harvested and prepared for Western blots as above.

Antifungal susceptibility tests
Minimal inhibitory concentrations (MICs) of fluconazole were determined using the broth microdilution method according to CLSI guideline M27-A3. Single colonies of freshly streaked C. albicans cells (< three days old) were scrapped off a YPD plate and diluted to 2 x 10 3 cells/ ml in YPD medium. Fluconazole drug concentrations ranged from 0.25 μg/ml to 128 μg/ml. One-hundred-microliter of 2-fold serial dilution of the drugs were added into wells of 96-well plates containing 100 μl of C. albicans. A plate reader (Tecan) was used to determine optical density at OD 600nm after 24 h and 48 h. MIC was defined as the lowest concentration resulting in no fungal growth. MIC 80 was defined as the concentration resulting in an inhibition of at least 80% of fungal growth. For analysis of susceptibility to various stressors, ten-fold serial dilutions of overnight-grown cultures starting from an optical density (OD 600nm ) of 0.5 were plated on control plates or plates supplemented with various compounds as indicated in the figure legends. Plates were incubated at 30˚C or 37˚C for 2 days and then photographed.

RNA-seq analysis
Wild type and rpn4Δ/Δ cells from overnight-grown cultures were diluted to an OD 600nm of 0.2 and grown in 20 ml of YPD medium at 37˚C with shaking at 200 rpm till log phase (OD 600nm of 1). Log phase cultures were treated with 3 μg/ml fluconazole or the solvent control DMSO (final concentration 0.06%) for 30 min at 37˚C with shaking at 200 rpm. Cultures (10 ml) were then centrifuged at 3000 rpm at 4˚C for 3 min, washed with water and pelleted into screw-cap tubes as '+FLC' or '-FLC' samples. Cell pellets were snap-frozen on dry ice and stored at -80˚C awaiting RNA extraction.
Total RNA from three independent biological repeats were extracted using the hot phenol method as we described previously [46]. Quant-seq libraries were sequenced with the Illumina Hiseq1500 platform according to the manufacturers' instructions. The raw sequencing data were processed as described previously [46,49]. The cut-off for genes to be included in the analysis was � 5 counts per million (CPM) in at least 2 samples. The RNA-seq dataset can be viewed at https://degust.erc.monash.edu/degust/compare.html?code= 7626de961da6ddac4911adffc3b6e7ca#/. The volcano plots in Fig 2A were Fig 2C shows the overlap between genes differentially expression in wild type and rpn4Δ/Δ cells by fluconazole (FDR < 0.05, log2 of 0.585, i.e. � 1.5 fold change). The gene ontology (GO) term analysis of biological process in differentially expressed genes was performed using the tools at the Candida Genome Database.
Genome sequencing was performed on wild type (YCAT 641, SN425), rpn4Δ/Δ clone x (YCAT 1099) and rpn4Δ/Δ clone y (YCAT 1100) strains (S1 Table). Strains were grown in standard culture conditions (growth for 14 h in YPD medium 30˚C) and genomic DNA was extracted via the phenol/chloroform/isoamyl alcohol (25:24:1) method. Whole genome sequencing (WGS) was performed by Deakin Genomics Centre on a MiSeq V3 machine. For the three strains, WGS libraries were prepared with Illumina DNA Prep with unique dual indexing, using an insert size of 300bp. To check for evidence of changes to chromosome copy number the WGS paired-end fastq files were uploaded to the YMAP analysis site [50]. The reference genome selected was SC5314 (ver. A21-s02-m08-r09) with the wild type strain (YCAT 641) selected as the parental strain for analysing rpn4Δ/Δ clone x and rpn4Δ/Δ clone y strains. The results of this analysis are shown in S5 Fig. Further to this, the RNA-seq data was analysed for evidence of chromosomal bias in gene expression. The sequence data was processed as described above for RNA-seq to produce gene-level counts per sample. These counts were TMM normalised [51] and converted to log CPM values before fitting a voom/limma [52] linear model to estimate log fold change per gene shown in boxplots in S6A and S6B Fig. A normal distribution per chromosome was fitted to these producing an estimated log fold change per chromosome which was used to test whether the chromosome showed evidence of biased gene expression.

PLOS PATHOGENS
Rpn4, the proteasome and azole drug tolerance S2 Table. Primers used in this study. (XLSX) S1 Movie. Live cell imaging of infected macrophages. Murine bone marrow-derived macrophages were infected with C. albicans (wild type, rpn4Δ/Δ clones x and y and their respective complemented strains. Hyphal formation in macrophages was viewed with brightfield live cell microscopy from 2-9 h post-infection. The scale bar represents 100 μm. Clips were generated using FIJI; brightness and contrast maxima were decreased to 12 000-15 000. (MP4)