Elevated catalase expression in a fungal pathogen is a double-edged sword of iron

Most fungal pathogens of humans display robust protective oxidative stress responses that contribute to their pathogenicity. The induction of enzymes that detoxify reactive oxygen species (ROS) is an essential component of these responses. We showed previously that ectopic expression of the heme-containing catalase enzyme in Candida albicans enhances resistance to oxidative stress, combinatorial oxidative plus cationic stress, and phagocytic killing. Clearly ectopic catalase expression confers fitness advantages in the presence of stress, and therefore in this study we tested whether it enhances fitness in the absence of stress. We addressed this using a set of congenic barcoded C. albicans strains that include doxycycline-conditional tetON-CAT1 expressors. We show that high basal catalase levels, rather than CAT1 induction following stress imposition, reduce ROS accumulation and cell death, thereby promoting resistance to acute peroxide or combinatorial stress. This conclusion is reinforced by our analyses of phenotypically diverse clinical isolates and the impact of stochastic variation in catalase expression upon stress resistance in genetically homogeneous C. albicans populations. Accordingly, cat1Δ cells are more sensitive to neutrophil killing. However, we find that catalase inactivation does not attenuate C. albicans virulence in mouse or invertebrate models of systemic candidiasis. Furthermore, our direct comparisons of fitness in vitro using isogenic barcoded CAT1, cat1Δ and tetON-CAT1 strains show that, while ectopic catalase expression confers a fitness advantage during peroxide stress, it confers a fitness defect in the absence of stress. This fitness defect is suppressed by iron supplementation. Also high basal catalase levels induce key iron assimilatory functions (CFL5, FET3, FRP1, FTR1). We conclude that while high basal catalase levels enhance peroxide stress resistance, they place pressure on iron homeostasis through an elevated cellular demand for iron, thereby reducing the fitness of C. albicans in iron-limiting tissues within the host.


Introduction
Of the circa three million fungal species that are thought to inhabit our planet [1], only a relatively small number have been reported to cause infections in humans. (About 400 species are described in the Atlas of Clinical Fungi [2].) Nevertheless, there is an increasing awareness that these fungal pathogens impose a major burden on human health worldwide [3]. These clinically important fungi generally share common features that promote colonization of their human host, such as the thermotolerance that permits growth at body temperatures. These common features include relatively robust stress responses, which mitigate against the stresses imposed by host immune defences [e.g. [4][5][6]. They also include the ability to scavenge essential micronutrients, such as iron, from their host [7][8][9][10].
Iron is an essential micronutrient that is required for the functionality of key ferroproteins and haem proteins. However, excess iron is toxic because ferrous ions promote the Fenton reaction which produces highly toxic hydroxyl radicals [11], and therefore host and pathogen alike must tightly regulate their acquisition, storage and release of iron. Consequently, the levels of free ion are vanishingly low in some host niches [12]. Furthermore, following infection the host activates the process of nutritional immunity in an effort to limit iron availability for the invading microbe [10,12]. Fungal pathogens respond to this iron limitation by down-regulating genes encoding iron-containing proteins and upregulating efficient iron scavenging mechanisms [13][14][15][16][17]. In Candida albicans this response includes the induction of genes encoding ferric reductases (e.g. CFL5, FRP1), high affinity iron permeases (e.g. FTR1) and proteins involved in iron assimilation (e.g. FET3) [15]. This response allows the fungus to counter the changes in iron homeostasis within the host that are triggered by systemic candidiasis [10].
Fungal pathogens activate oxidative stress responses when they come in contact with the host [18][19][20][21][22], and these responses promote resistance to phagocytic attack and fungal virulence [5,[23][24][25][26]. In an attempt to clear invading fungal pathogens, host neutrophils and macrophages phagocytose the fungal cells and subject them to a battery of reactive oxygen species (ROS) that damage proteins, DNA and lipids, and can induce programmed cell death [27]. The impact of ROS is augmented when combined with a cationic stress, and this synergistic impact of combinatorial oxidative and cationic stresses is thought to contribute to the potency of human neutrophils [28,29]. C. albicans cells respond to oxidative stress by inducing functions that detoxify the ROS, repair the oxidative damage, synthesize antioxidants and restore redox homeostasis. This includes the induction of genes encoding catalase (CAT1), superoxide dismutases (SOD), glutathione peroxidases (GPX) and components of the glutathione/glutaredoxin (GSH1, TTR1) and thioredoxin (TSA1, TRX1, TRR1) systems [6,[30][31][32]. In particular, CAT1 mRNA levels are strongly induced by oxidative stress [30,33]. However, C. albicans cells are unable to activate a normal transcriptional response to oxidative stress when subjected to combinatorial oxidative plus cationic stress or acute peroxide stress, and this contributes to the lethality of these types of stress [28,29].
Catalase (Cat1) plays a major role in protecting C. albicans against peroxide stress [28,29]. This iron-requiring enzyme, which has been well-characterised structurally [34], belongs to a superfamily of heme peroxidases and catalases that are conserved across bacteria, plants, fungi and animals [35]. Catalase catalyses the conversion of hydrogen peroxide (H 2 O 2 ) to water. C. albicans cells rapidly detoxify extracellular H 2 O 2 following exposure to an acute peroxide stress, and this detoxification is mainly dependent on catalase (CAT1) [28].
We showed previously that ectopic expression of catalase using the ACT1 promoter (ACT1 p -CAT1) protected C. albicans from acute oxidative and combinatorial stresses [28]. More recently, Jesus Pla's group has confirmed that catalase overexpression protects C. albicans against peroxide stress [36]. They also demonstrated that high catalase levels provide protection against antifungal drugs. These observations raise an interesting conundrum: if catalase overexpression confers effects that might be expected to promote host colonisation, why has C. albicans not evolved to express high basal levels of catalase? We address this in this study. We show that while high basal catalase levels enhance the fitness of C. albicans in the presence of oxidative and combinatorial stresses, these high catalase levels reduce fitness in the absence of stress. We also reveal the molecular basis for this fitness defect. Our observations suggest a partial explanation for the lack of emergence of catalase overexpression during the evolution of this major fungal pathogen. We also show that, in contrast to the prevailing view [23], the virulence of C. albicans is not compromised by catalase inactivation.

High basal catalase levels protect against oxidative and combinatorial stresses
We demonstrated previously that ectopic expression of catalase from an ACT1 promoter-CAT1 fusion (ACT1 p -CAT1) reproducibly protected C. albicans against acute peroxide stress (5 mM H 2 O 2 ) and combinatorial stress (5 mM H 2 O 2 plus 1 M NaCl) [28]. Subsequently we noted that the stress resistance of ACT1 p -CAT1 cells declined over time (S1 Fig). Therefore, we constructed new C. albicans strains in which catalase expression is regulated by the doxycycline conditional tetON promoter [37][38][39]. Control strains were made by transforming congenic wild-type (CAT1) and catalase null strains (cat1Δ) with empty tetON vectors. Test strains were made by integrating a tetON-CAT1 plasmid into the genome of the cat1Δ null mutant. We refer to these strains, which all have the same genetic background (Materials and Methods; S1 Table), as wild-type (CAT1), null (cat1Δ) and tetON-CAT1 strains, respectively. Three isolates were generated for each strain type. For each strain type, the isolates displayed similar stress phenotypes (below).
First we tested Cat1 expression levels in wild-type (CAT1), null (cat1Δ) and tetON-CAT1 cells. Catalase levels were induced in response to oxidative stress in wild-type (CAT1) cells, and were undetectable in cat1Δ cells (Fig 1). Catalase levels in these strains were not affected by doxycycline addition. In contrast, catalase levels were strongly induced by doxycycline in tetON-CAT1 cells (red bars, Fig 1). Significantly, wild-type cells express significant basal levels of catalase in the absence of stress (Fig 1), as we reported previously [33]. Catalase levels in doxycycline-treated tetON-CAT1 cells were higher than these basal levels (Fig 1).
We then compared the stress resistance of wild-type, null and tetON-CAT1 cells (Fig 2A &  S1 Fig). As expected [23,28,40], wild-type (CAT1) cells displayed modest resistance to an oxidative stress (H 2 O 2 ) and a combinatorial stress (H 2 O 2 plus NaCl), whereas the null mutant (cat1Δ) was sensitive to both types of stress. These phenotypes were not affected by the presence or absence of doxycycline (Fig 2A). In the absence of doxycycline, the tetON-CAT1 strains were sensitive to both oxidative and combinatorial stress, reflecting their null cat1Δ background. When these strains were pre-grown with doxycycline, they displayed enhanced oxidative and combinatorial stress resistance (Fig 2A). This reinforces our earlier conclusion [28] that elevated basal CAT1 expression levels protect C. albicans cells against a sudden and acute oxidative or combinatorial stress.
Interestingly, the tetON-CAT1 strains were sensitive to both oxidative and combinatorial stress when pre-grown in the absence of doxycycline and the inducer was only provided when the stress was imposed (Fig 2A). We then measured the impact of individual and combinatorial oxidative (H 2 O 2 ) and cationic (NaCl) stresses upon cell death by cytometric analysis of propidium iodide (PI) stained doxycycline-grown cell populations (Fig 2B). Relative to the control wild-type strain, cat1Δ null cells were more sensitive, and tetON-CAT1 cells were more resistant to these oxidative and combinatorial stresses. Compared to the control wild type cells, doxycycline-treated tetON-CAT1 cells displayed 9-fold less cell death following exposure to the oxidative stress, and 2.5-fold less death after the combinatorial stress ( Fig 2B).
This correlated with a reduction in internal ROS accumulation following stress imposition by tetON-CAT1 cells relative to the wild-type and cat1Δ cells (Fig 2C). The accumulation of intracellular ROS was 2.6-fold lower in doxycycline-treated tetON-CAT1 cells after the peroxide stress, and 1.5-fold lower following the combinatorial stress, compared to the wild type control ( Fig 2C). Taken together, our data indicate that cells with low catalase levels at the point of stress imposition are more sensitive to peroxide than cells with high catalase levels. This suggests if catalase levels are low at the point of stress imposition, the dynamics of catalase induction are too slow to permit the normally rapid clearance of peroxide [28] and to prevent ROS-mediated cell death [27]. The data indicate that C. albicans cells require high basal levels of catalase at the time of stress imposition if they are to survive an acute oxidative or combinatorial stress.
Oxidative stress resistance and catalase levels in clinical isolates C. albicans clinical isolates display a high degree of natural variation [41,42]. We exploited this to select strains that display relatively low levels of oxidative stress resistance. A diverse range of C. albicans clinical isolates (65 in total) from different epidemiological clades and from different patient colonisation sites were subjected to a robotic screen in which they were plated on YPD containing different peroxide concentrations ( Fig 3A). All of the isolates tested displayed a degree of resistance to this stress, showing some growth at 3.2 mM H 2 O 2 . However, some isolates were more resistant to peroxide, displaying robust growth at 6.4 mM H 2 O 2 , whereas sensitive strains were unable to grow at this H 2 O 2 concentration. We selected a subset of four sensitive isolates and three resistant isolates (which included SC5314, the clinical isolate from which most laboratory strains are derived), and compared the basal CAT1 expression levels in these isolates to a standard laboratory strain (CAI4 containing CIp10 (URA3)). Basal CAT1 mRNA levels were lower in the oxidative stress sensitive isolates tested (Fig 3B), and furthermore, the basal levels of the enzyme were also lower in these isolates ( Fig 3C). These data are consistent with the idea that elevated basal catalase levels promote oxidative stress resistance in C. albicans.
Population heterogeneity promotes C. albicans survival following oxidative stress Next we examined how a subset of cells within an apparently homogeneous population of C. albicans cells can survive an acute oxidative stress [28,36,43]. Based on the above observations, we reasoned that this might be partly explained by stochastic variation in basal catalase levels between individual cells in such a population. To test for potential population heterogeneity in basal catalase levels we generated a strain in which both CAT1 alleles were tagged with GFP (CAT1-GFP/ CAT1-GFP) to express a Cat1-GFP fusion protein. Western blotting revealed a Cat1-GFP protein of the expected mass in these cells (approximately 80 kDa: Fig 4A), and the GFP fluorescence was located in punctate spots (Fig 4B), consistent with the peroxisomal localisation of catalase in C. albicans [44]. We then compared the oxidative stress resistance of the CAT1-GFP strain with congenic control wild-type (CAT1/CAT1), heterozygous (CAT1/ cat1Δ) Catalase activities were measured in protein extracts from mid-exponential C. albicans cultures containing 0 or 20 μM doxycycline (-or + Dox, respectively): cat1Δ, Ca2089; wild-type, WT, Ca2084; red, tetON-CAT1 isolates, Ca2040, Ca2043, Ca2046 (S1 Table). Immediately before harvesting, wild-type and cat1Δ cultures were exposed to 0 or 5 mM H 2 O 2 for one hour. Means and standard deviations from three independent replicate experiments are shown, and the data were analysed using one-way ANOVA with Tukey's post-hoc test: *, p 0.05; **, p 0.01; ***, p 0.001; ****, p 0.0001. and null (cat1Δ/cat1Δ) strains. The CAT1-GFP strain was as resistant to oxidative stress as the wild-type control (Fig 4C), indicating that the CAT1-GFP alleles are functional.
We then examined the basal levels of GFP fluorescence in unstressed populations of exponentially growing C. albicans CAT1-GFP cells by flow cytometry. As predicted, these genetically homogeneous cell populations displayed heterogeneity with respect to their Cat1-GFP expression levels ( Fig 4D & S2 Fig). Using flow cytometry, we selected cells of similar size, sorted cells that display relatively low levels of Cat1-GFP from those expressing high levels ( Fig  4D & S2 Fig), and then plated them onto media containing a range of H 2 O 2 concentrations. Cells expressing relatively high levels of Cat1-GFP were more resistant to peroxide stress ( Fig  4E). When an analogous experiment was performed with cells expressing a control gene (ACT1-GFP), stochastic differences in ACT1-GFP expression did not affect oxidative stress resistance (S3 Fig). These observations reinforce our conclusion that high basal levels of catalase promote oxidative stress resistance. Furthermore, this confirms that C. albicans cell populations display stochastic variation in their basal CAT1 expression levels, and that this contributes to the survival of a subset of C. albicans cells following an acute oxidative stress.

Impact of catalase levels on host colonisation during systemic infection
We tested whether high basal catalase levels affect the ability of C. albicans to colonise different tissues during systemic infection. At first we reasoned that the elevated oxidative stress resistance conferred by high basal catalase levels (above) might enhance host colonisation. To test this we compared directly the three isolates for wild-type (CAT1), null (cat1Δ) and tetON-CAT1 strains (nine in total) using a barcode sequencing (barseq) strategy. The C. albicans strains were pre-grown separately in the presence or absence of doxycycline. Approximately equal amounts of the nine doxycycline-treated strains were mixed and used to induce disseminated candidiasis in doxycycline-treated mice. In parallel, the nine untreated control C. albicans strains were mixed and used to infect untreated mice. Mice from each group were culled after four days, and the fungal cells harvested from their kidneys, livers, spleens and brains. Barseq was then performed on genomic DNA from these fungal populations to determine the relative proportion of each C. albicans strain in each tissue. We observed significant differences between the wild-type (CAT1) and tetON-CAT1 strains in their ability to colonise certain tissues ( Fig 5). Doxycycline-treated tetON-CAT1 cells were less able to colonise the kidney and  Table). (C) Basal catalase levels in the same strains were assayed under the same growth conditions as (B). Means and standard deviation from three replicates are presented, and the difference between the sets of sensitive and resistance strains compared statistically: ****, p 0.0001.
https://doi.org/10.1371/journal.ppat.1006405.g003 brain than the control untreated tetON-CAT1 cells, but this was not the case in the liver and spleen. This effect was observed for tetON-CAT1-1 cells, but not for the other two tetON-CAT1 isolates (4 and 10: S1 Table). This correlated with a reduction in catalase levels in these isolates (S4A Fig) and a corresponding loss of phenotype (S4B Fig). Therefore, like ACT1 p -CAT1 cells (above; S1 Fig), isolates 4 and 10 appeared to have lost their phenotype over time. Taken together, our data indicate that, contrary to our initial prediction, high basal catalase expression levels appear to compromise, rather than enhance, the ability of C. albicans to colonise certain tissues.
To our surprise, we did not observe any significant differences between the wild-type (CAT1) and null (cat1Δ) strains in their ability to colonise the host ( Fig 5). All of the wild-type and null isolates displayed similar levels of colonisation. This indicated that cells lacking catalase can infect the host-a conclusion that contrasts with the prevailing view that C. albicans cat1Δ null cells display attenuated virulence [23,40]. We reasoned that cat1Δ cells might be  Table). able to colonise host tissues if they are co-infected with CAT1 and tetON-CAT1 cells. For example, cat1Δ null cells might be rescued via a "cheater" or "bystander" effect [45,46], whereby catalase expressing cells protect the null mutant against local peroxide stress.
We tested this by comparing the virulence of our wild-type (CAT1) and null (cat1Δ) strains separately in the three-day murine model of systemic candidiasis [47]. We observed no significant difference between the wild-type or mutant strains with respect to their fungal burdens in the kidneys, and the strains induced similar levels of weight loss in mice, yielding similar outcome scores that displayed no significant difference (Fig 6A). This observation reinforced the idea that inactivating CAT1 does not attenuate the virulence of C. albicans.
Wysong and co-workers observed a virulence defect for cat1Δ cells using a long-term mouse model of systemic candidiasis [23]. Therefore, it seemed possible that our short-term  Table). Mice that received the pool of nine C. albicans strains pre-grown with doxycycline were treated with doxycycline, whereas mice that were infected with the pool of C. albicans strains pre-grown without doxycycline did not (Materials and Methods). After four days, mice were killed by cervical dislocation and the kidneys, brains, livers and spleens harvested. The fungal cells infecting these organs were grown on YPD plates, genomic DNA isolated from each fungal population, and the abundance of each barcode in each fungal population was quantified by barseq. The Relative Abundance of each barcode in each tissue from each mouse was calculated relative to the total number of barcode reads for that specific sample, and then normalised against the proportion for that barcode in the pool of nine C. albicans strains used to infect that mouse. Each symbol represents data for a single mouse. The pairwise comparisons indicated were analysed using the Students ttest: *, p 0.05; **, p 0.01. Data are presented for three of the nine barcoded C. albicans strains: CAT1 strain 23, grey and black; cat1Δ strain 28, pink and red; tetON-CAT1 strain 1, pale blue and blue; no doxycycline, grey, pink, pale blue; 20 μM doxycycline, black, red, blue. These wild type and cat1Δ strains were representative of the other isolates. However, tetON-CAT1-1 behaved differently from the other tetON-CAT1 isolates, which were shown subsequently to have lost their phenotype (below). and their long-term model of systemic infection might yield differing outcomes for C. albicans cat1Δ cells. To test this we re-examined the virulence of our wild-type (CAT1) and null (cat1Δ) strains in mice over 14 days. (We were unable to access the strains used by Wysong and coworkers [23]. Hence we could not perform a direct comparison with their mutant.) No major difference in the virulence of wild-type and cat1Δ cells was observed using a long term infection model (p = 0.074: Fig 6B). We also compared our wild-type (CAT1), null (cat1Δ) and tetON-CAT1 strains in Galleria mellonella, observing no significant difference in their virulence in this invertebrate model of systemic candidiasis (p = 0.68: Fig 6C).
The cat1Δ mutant generated by Wysong and co-workers had the URA3 marker inserted at the cat1 locus (cat1::URA3) [23]. In contrast, in our cat1Δ mutant URA3 was reintroduced at the RPS1 locus using the CIp10 plasmid backbone [48]. After the study of Wysong and coworkers was published [23], URA3 position effects were found to influence C. albicans virulence, and reinsertion of URA3 at RPS1 using CIp10 was shown to overcome these effects [49]. We conclude that CAT1 inactivation does not significantly attenuate the virulence of C. albicans.

Catalase promotes resistance to neutrophil killing
It has been reported that catalase null mutants do not display significantly higher sensitivities to neutrophil killing [5]. Once again, these experiments were performed with a cat1Δ null mutant in which URA3 was integrated at the CAT1 locus (cta1Δ::loxP-URA3-loxP: [5]). Therefore, in light of our findings (above), we retested neutrophil killing using our new cat1Δ strain in which URA3 is integrated at the RPS1 locus. We tested the strains separately to exclude potential cheater effects [45,46]. We observed that, following exposure to human neutrophils, our new cat1Δ strain displayed significantly reduced survival compared to the congenic wild-type control (Fig 7). This strengthens the observation of Miramon and co-workers, who reported a slight difference between cat1Δ and CAT1 cells that was not statistically significant [5]. Furthermore, we also observed a statistically significant difference in neutrophil killing between tetON-CAT1 cells that were pre-grown in the presence or absence of doxycycline (Fig 7). These data indicate that catalase promotes the resistance of C. albicans against neutrophil attack. We note that elevated basal levels of catalase did not enhance the resistance of C. albicans to neutrophil killing in our hands (Fig 7: compare wild type and doxycycline-treated tetON-CAT1 cells).
Elevated catalase levels affect the ability of C. albicans to compete in vitro C. albicans cat1Δ cells are clearly sensitive to oxidative stress (Fig 2). However, in mixed populations they could conceivably be rescued by catalase expressing cells. Therefore, we tested whether cat1Δ cells act as cheaters by examining their fitness in mixed cultures alongside wildtype (CAT1) and tetON-CAT1 cells. The three barcoded for the null mutant, wild-type and tetON-CAT1 strains were pre-grown separately in the presence of doxycycline, mixed in approximately equal proportions, and then used to inoculate YPD cultures containing doxycycline. A parallel mixture of untreated barcoded strains was also prepared, and this untreated mixture used to inoculate YPD cultures without doxycycline. The relative fitness of each strain was then compared in the presence or absence of oxidative stress (5 mM H 2 O 2 ), by comparing the relative abundance of each barcode over time in each culture by barseq. With one notable exception (discussed below), the three isolates for each strain type displayed similar behaviours (Fig 8).
In the absence of doxycycline and stress, the relative abundance of the wild-type (CAT1), null (cat1Δ) and tetON-CAT1 strains did not change significantly over the twelve hour period examined (Fig 8A). In contrast, in the absence of doxycycline but in the presence of stress, the abundance of cat1Δ and tetON-CAT1 cells rapidly declined in the population and these strains were rapidly outcompeted by the wild-type CAT1 strains (Fig 8B). The comparable behaviour Impact of catalase on resistance to neutrophil killing. C. albicans wild-type (CAT1, black, Ca2084), cat1Δ (red, Ca Ca2089) and tetON-CAT1 cells (Ca2038) pre-grown with 0 or 20 with μM doxycycline (pale blue and blue, respectively) were exposed to human neutrophils for 2 h, and then fungal survival assayed. Each data point represents the mean for three replicates from one healthy donor. The data were analysed using one-way ANOVA with Tukey's post-hoc test: *, p 0.05; **, p 0.01. https://doi.org/10.1371/journal.ppat.1006405.g007 Catalase expression affects iron demand and fitness for the cat1Δ the tetON-CAT1 cells under these conditions was entirely consistent with the negligible catalase levels in tetON-CAT1 cells without doxycycline induction (Fig 1). These data strongly reinforce the view that catalase is vital for peroxide stress resistance in C. albicans [5,23,33,36,40]. Our data also show that cat1Δ cells do not act as cheaters: they are not rescued by catalase expressing cells under peroxide stress conditions (Fig 8B).
In the presence of doxycycline in the presence of stress, the tetON-CAT1 cells rapidly outcompeted the null (cat1Δ) cells (Fig 8D). This again highlighted the peroxide sensitivity of cat1Δ cells. Significantly, the tetON-CAT1 cells also out-competed wild-type (CAT1) cells ( Fig  8D), confirming directly that ectopic catalase expression enhances oxidative stress resistance (Fig 2) [28,36]. Therefore, elevated basal catalase levels increase the fitness of C. albicans cells in the presence of peroxide stress.
Interestingly, in the presence of doxycycline but in the absence of peroxide stress, there was a decrease in the abundance of tetON-CAT1-01 cells in the population over the twelve hour The nine barcoded C. albicans wild-type, cat1Δ and tetON-CAT1 strains (S1 Table) were pre-grown separately in YPD containing 0 or 20 μM doxycycline. The cells grown in the presence of doxycycline were then mixed in roughly equal proportions and used to inoculate fresh YPD cultures containing doxycycline (plus Dox). In parallel, the cells grown in the absence of doxycycline were mixed and used to inoculate fresh YPD cultures lacking doxycycline (no Dox). Parallel cultures contained 0 or 5 mM H 2 O 2 (no stress and plus stress, respectively). Cells were harvested at various time points over a 12 h period, and barseq was performed on genomic DNA extracted from each sample. The relative abundance of each barcode in each sample was then calculated relative to its starting abundance. Data are shown for one barcoded strain of each type (means and standard deviation from three replicate measurements): wild-type (CAT1), black, Ca2084; cat1Δ, red, Ca2089; tetON-CAT1, blue, Ca2038. time-course, relative to the wild-type (CAT1) and null (cat1Δ) cells (Fig 8C). This suggested that ectopic CAT1 expression might render C. albicans cells less fit in the absence of stress.

Ectopic catalase expression confers a fitness defect in the absence of stress that is suppressed by iron supplementation
Doxycycline-treated C. albicans tetON-CAT1 cells appeared to display a fitness defect in the absence of stress (Fig 8C). We tested this further by examining biomass formation on YPD (final OD 600 ) (Fig 9A). All of the strains displayed similar growth in the absence of doxycycline, and the wild-type (CAT1) controls remained unaffected by doxycycline. However, the growth of tetON-CAT1 cells decreased in the presence of doxycycline, reinforcing the view that elevated catalase levels reduce fitness in the absence of stress.

Fig 9. In the absence of stress, elevated catalase levels impose a fitness defect on C. albicans that is suppressed by iron supplementation. (A)
The growth of C. albicans tetON-CAT1 isolates (1, 4, 10) was monitored (OD 600 ) in YPD containing 0 or 20 μM doxycycline (Ca2040, Ca2043, Ca2046; S1 Table). The impact of tetON-CAT1 induction upon fitness was assayed by subtracting the OD after growth in the absence of doxycycline (OD -Dox ) from the OD in the presence of doxycycline (OD +Dox ). Data represent the means and standard deviations from three independent experiments. (B) The impact of iron on the fitness of C. albicans tetON-CAT1 strain 1 (Ca2040) was measured in YPD cultures containing 0 or 20 μM doxycycline plus different concentrations of FeCl 3 . Data represent the means and standard deviations from three independent experiments. (C) The effect of tetON-CAT1 expression on genes involved in iron assimilation and homeostasis was assessed by qRT-PCR of specific transcripts (relative to the ACT1 mRNA internal control) during growth of C. albicans tetON-CAT1 isolate 1 (Ca2040) in YPD containing 0 or 20 μM doxycycline. The data, which represent the means and standard deviations from three replicate measurements, are normalised relative to the corresponding transcript level in wild type cells grown with doxycycline. Catalase is a ferroprotein [34] expressed at relatively high basal levels in C. albicans (approximately 1.5 x 10 5 molecules per cell [33]). In bacteria, catalase overexpression has been reported to affect the requirement for iron [50]. Therefore, we reasoned that the fitness defect conferred by high basal catalase levels in C. albicans might be mediated by an elevated cellular demand for iron. Hence, we tested whether iron supplementation can suppress this fitness defect. Growth of tetON-CAT1 cells was measured in YPD containing doxycycline supplemented with different concentrations of ferric ions (Fig 9B). These data indicate that the fitness defect caused by ectopic catalase expression can be completely suppressed by iron supplementation. This suppression was due to the improved growth of doxycycline-treated tetON-CAT1 cells in the presence of iron (S5A Fig). We also showed that iron supplementation suppresses the reduced fitness of doxycycline-treated tetON-CAT1 cells in direct competition experiments with wild type (CAT1) cells (S5B Fig).
These observations suggested that high basal catalase expression increases the cellular demand for iron in C. albicans. To test this further we examined the impact of ectopic catalase expression upon key genes involved in iron assimilation and homeostasis: CFL5 (encoding a ferric reductase that is induced in low iron), FET3 (encoding a copper oxidase that is required for growth in low iron), FRP1 (encoding a ferric reductase that is induced by iron chelation) and FTR1 (encoding a high-affinity iron permease that is required for growth in low iron). All of these genes are targets of the iron-responsive transcriptional activator Sef1 [15]. CFL5, FET3, FRP1 and FTR1 transcript levels were measured relative to the ACT1 mRNA internal control in tetON-CAT1 cells grown in the presence and absence of doxycycline. Their levels were then normalised against those in doxycycline-treated wild type (CAT1) cells to exclude any potential effects of this treatment on these transcripts [51]. All four iron-responsive transcripts were strongly induced following tetON-CAT1 induction (Fig 9C). Taken together, the data indicate that high basal catalase levels increase the requirement for iron in C. albicans.

Discussion
This study has important implications for the impact of the key peroxide detoxifying enzyme, catalase, upon the stress resistance and virulence of the major fungal pathogen, C. albicans. Firstly, our analyses of new cat1Δ null mutants, in which potential URA3 position effects have been circumvented [49], have reinforced the view that catalase is essential for normal levels of oxidative and combinatorial stress resistance in C. albicans (Figs 2 & 8). They also show that catalase contributes to the resistance of this pathogenic fungus against neutrophil killing ( Fig  7). However, our most surprising finding was that, in contrast to the generally held view [23,40], catalase is not essential for the virulence of C. albicans, at least in models of disseminated candidiasis. This unexpected finding is supported by virulence assays in both short term and long term murine models of systemic infection, and in an accepted invertebrate model of systemic infection (Fig 6). This view is further reinforced by our in vivo competition experiments, in which the cat1Δ null mutant competed effectively against wild-type and catalase overexpressing strains for colonisation of the kidney, liver, spleen and brain (Fig 5). We suggest that the attenuated virulence of the cat1Δ mutants reported previously [23,40] might be explained by URA3 position effects in these strains [49].
Why might catalase be important for oxidative stress resistance and yet apparently not required for systemic infection? The sensitivity of cat1Δ cells to neutrophil killing (Fig 7) does indicate that protection against peroxide is required in certain contexts in vivo. Therefore, this lack of cat1Δ virulence defect probably reflects the multifactorial nature of virulence phenotypes, as well as the nature of the systemic infection models often used to examine virulence. In these models sufficient fungal doses are applied to overcome immediate clearance by circulating phagocytes [47]. Furthermore, few of the fungal cells colonising the kidney appear to be exposed to oxidative stress [31].
Secondly, our data indicate that high basal levels of catalase promote the resistance of C. albicans to peroxide and combinatorial stress (Fig 2). These data reaffirm previous reports that elevated catalase expression promotes peroxide resistance [28,36]. Significantly, our data indicate that this phenotype is dependent on high basal levels of catalase at the point of stress imposition, rather than CAT1 induction in response to stress. Three independent observations support this view. (A) tetON-CAT1 cells are only protected against peroxide or combinatorial stress if these cells are pre-treated with doxycycline, not if doxycycline is only provided at the same time as the stress (Fig 2). (B) Clinical isolates that are relatively resistant to oxidative stress tend to express catalase at relatively high levels (Fig 3). (C) Unstressed C. albicans cell populations display heterogeneity in Cat1-GFP levels, and those cells that express more Cat1-GFP are less susceptible to killing by oxidative stress (Fig 4). Hydrogen peroxide is normally rapidly detoxified by wild-type C. albicans cells (within 60 minutes) in a catalase-dependent fashion [28]. Elevated basal levels of catalase presumably enhance cellular protection by accelerating the clearance of this reactive oxygen species. The heterogeneity in catalase expression within C. albicans populations, which might arise via stochastic differences between cells [52][53][54], appears to account, in large part, for the ability of a subset of C. albicans cells to survive an acute oxidative stress. This would appear to represent the first example in C. albicans of the kind of "bet-hedging" strategies that have been observed in bacterial and S. cerevisiae populations [55,56]. Furthermore, these observations are entirely consistent with the well-established observation that an adaptive response to a small dose of a particular stress can transiently endow yeasts with resistance to a subsequent acute dose of the same stress by inducing the expression of key stress protective functions. This observation has been reported for heat shock, osmotic and oxidative stress in S. cerevisiae for example [57,58], and has been extended to other yeasts including C. albicans [43,59,60].
Thirdly, our data provide key insights into the impact of catalase levels on the virulence of C. albicans. In our hands, direct competition assays suggested that elevated catalase levels might affect C. albicans colonisation of the kidney and brain (Fig 5). This is consistent with a parallel study which reported that catalase overexpression attenuates the virulence of C. albicans [36]. Roman and co-workers described this as "a most unexpected result" given that catalase overexpression enhances oxidative stress resistance. They speculate that this might have arisen via some alteration in fitness, which they were unable to detect in vitro, but which might interfere with activation of the Hog1 and Mpk1 MAP kinases [36]. In this study we show clearly in direct competition assays that elevated basal catalase levels attenuate the fitness of C. albicans in the absence of stress (Fig 8). We conclude that catalase overexpression confers a selective disadvantage in C. albicans in the absence of stress.
Fourthly, we have identified a major cause of this fitness defect. High basal catalase levels increase the cellular requirement for iron in C. albicans. We present two key observations that support this. (i) The fitness defect is suppressed by iron supplementation (Fig 9B and S4 Fig). This effect, which has also been observed in bacteria [50], is probably mediated by the depletion of intracellular iron through high level expression of an abundant heme-requiring enzyme. (ii) Ectopic catalase expression induces the expression of iron-responsive genes that play key roles in iron scavenging and homeostasis: e.g . CFL5, FET3, FRP1 and FTR1 (Fig 9C). Therefore, the demand for iron and catalase expression are intimately linked in C. albicans. Both modulate the accumulation of intracellular ROS. Iron stimulates CAT1 expression in C. albicans [16,61]. This increase in catalase affects iron demand and homeostasis (Fig 9B & 9C) and also enhances the detoxification of hydrogen peroxide, thereby decreasing the production of highly toxic hydroxyl radicals via the iron-dependent Fenton reaction [11]. Parallels exist in S. cerevisiae, where heterogeneity in superoxide dismutase (SOD1) gene expression affects the fitness of individual cells in the presence of copper [62].
The impact of catalase levels on the requirement for iron is likely to have a profound effect on C. albicans pathogenicity because iron homeostasis is tightly regulated during infection [10,15] and efficient iron assimilation is essential for colonisation of iron limiting niches in the mammalian host [7]. It would appear significant, therefore, that we observed reduced colonisation for catalase overexpressing cells in the kidney and brain, but not in the iron-rich liver and spleen (Fig 5).
In conclusion, elevated basal catalase levels appear to be a double-edged sword whereby they protect C. albicans against oxidative and combinatorial stresses imposed by the host while increasing the pathogen's demand for an essential, but limiting micronutrient in the host. This double-edged sword would appear to account for the apparently counterintuitive observation that catalase overexpression in C. albicans decreases host colonisation in some tissues [36]. It also helps to explain why C. albicans has not evolved to express the high levels of catalase that would protect it from phagocytic killing [28,36].

Strains, growth conditions and treatments
The strains used in this study are listed in S1 Table. C. albicans was routinely grown at 30˚C, 200 rpm in YPD (2% dextrose, 2% mycological peptone, 1% yeast extract) containing 20 μg/ml doxycycline (Dox) when required. On the day of an experiment, overnight cultures were diluted into fresh YPD to an OD 600 of 0.2, and incubated at 30˚C at 200 rpm until they reached an OD 600 of 0.8, whereupon they were subjected to the appropriate treatment and analysed. Plates were incubated for 48 h at 30˚C.
Osmotic stress was applied with 1 M NaCl and oxidative stress was applied with H 2 O 2 at the specified concentration. Combinatorial stress was imposed using 1 M NaCl plus 5 mM H 2 O 2 as described previously [28,63].
Robotic plating was performed using a Singer RoToR robot (Singer Instruments, Watchet, UK). Fitness was assayed by monitoring growth in microtitre plates at OD 600 every 20 min for 48 h, and data from independent triplicate experiments were analysed.

Barcode sequencing
To quantify the relative concentration of each barcoded strain in mixed populations of tetON strains, genomic DNA was prepared from the populations by phenol: chloroform extraction method [66]. A 60 bp region carrying the barcodes (S3 Table) was amplified with common primers (S2 Table)  To analyse the barseq data, a wrapper script was coded over the open source BBDuk tool (BBMap suite version 35.43 [67]). The wrapper visits each sample directory and runs the 3rd-party bbduk.sh script over each of the compressed read 1 and read 2 FASTQ files, generating corresponding FASTQ output files for the "matched" and "not-matched" reads for each barcode. The wrapper then computes the total number of reads for each barcode and its abundance relative to the total number of barcode reads. The barseq data are presented as the relative abundance of a barcode normalised to its starting concentration in the population. Means and standard deviations from three replicate measurements are presented.

Transcript levels
RNA was extracted from C. albicans cells using the Zymo Research YesStar RNA Kit (Cambridge Bioscience, Cambridge, UK). cDNA was prepared using SuperScript II reverse transcriptase from Invitrogen (Fisher Scientific, Loughborough, UK), and qRT-PCR was performed with a Roche Light Cycler 480 II using the primers described in S2 Table. Transcript levels were measured in triplicate, expressed relative to the internal ACT1 mRNA control [28], and then normalised against the levels in doxycycline-treated wild type (CAT1) cells to exclude potential effects of doxycycline on these transcripts.

Catalase assays
C. albicans cells grown in YPD containing 0 or 20 μg/μl Dox were subjected to no stress or one hour of 5 mM H 2 O 2 , protein extracts prepared, and catalase activities measured using the BioAssay Systems EnzyChrom catalase assay kit (Universal Biologicals Ltd., Cambridge, UK), according to the manufacturer's instructions [28]. Assays were performed in triplicate. CAT1-GFP cells were visualized using a DeltaVision Core microscope (Applied Precision, Issaquah, WA). Western blotting was performed as described previously [68].

Cell viability and ROS accumulation
Cell viability was assayed by measuring colony forming units (CFU) on YPD plates and by propidium iodide (PI) staining and flow cytometry on a BD LSR II, as described previously [28,63].
Intracellular ROS accumulation was measured by staining the cells with 20 μM dihydroethidium for one hour in darkness, at 30˚C and 200 rpm, and then analysed using a BD LSR II flow cytometer. Data were analysed using Flowjo software version 10.0.8.

Neutrophil killing assays
Blood from healthy donors was obtained according to the protocol approved by the University of Aberdeen College Ethics Review Board (Application number-CERB/2012/11/676). Polymorphonuclear (PMN) cells, or neutrophils, were isolated from this blood using Histopaque-1119 and Histopaque-1077 (Sigma Aldrich) as described previously [28]. C. albicans cells pregrown with 20 μg/ml Dox were incubated with PMNs (1:10 ratio of yeasts to neutrophils) for 2 h in RPMI 1640 containing 10% heat inactivated foetal bovine serum. After incubation the PMNs were treated with 0.25% sodium docecyl sulphate and DNase I and yeast survival determined by assaying CFU. Data from eight healthy donors are presented with their means and standard deviation.

Virulence assays
The virulence of C. albicans wild type and cat1Δ cells were measured in a short term murine model of systemic candidiasis [47]. Strains were pre-grown in YPD and injected intravenously (4 x 10 4 CFU/g body weight) into the lateral tail vein of 6-10 week old female BALB/c mice (Envigo, UK). Mice were randomly assigned to cages (n = 6 per group) and inocula were randomly assigned to cages. Infections were allowed to proceed for 4 days whereupon the mice were humanely culled by cervical dislocation and fungal burdens (CFU/g) determined in the kidneys. Fungal burden and weight loss were used as measures of virulence [47].
The virulence of C. albicans wild type and cat1Δ strains were also tested in a longer term mouse infection model. Again, C. albicans cells were injected into the tail veins of 6-10 week old female BALB/c mice (3 x 10 4 CFU/g body weight). Once again, the mice were randomly assigned to cages (n = 8 per group) and inocula were assigned randomly to cages. The mice were monitored and weighed daily, and were humanely culled when they had lost 20% of their body weight and death recorded as having occurred on the following day. Experiments were continued for a maximum of 14 days, when all surviving mice were culled and analysed. The data are presented as Kaplan-Meier survival curves (log rank tests).
To directly compare the colonisation of C. albicans tetON strains in the mouse model of systemic candidiasis, the strains were pre-grown in YPD containing 0 or 20 μg/ml doxycycline and injected into the tail vein of 6-10 week old female BALB/c mice (4 x10 4 CFU/g body weight: n = 6 mice per group). Mice were gavaged with 100 μl of 0 or 40 mg/ml doxycycline. Infections were allowed to proceed for up to 4 days. Mice were culled, their kidneys, spleen, liver and brain removed and homogenized in 500 μl saline, and the entire sample from each organ plated onto YPD. The fungal colonies from each individual organ were then pooled, and genomic DNA prepared for barseq (above).
The virulence of C. albicans strains was also evaluated using the invertebrate Galleria mellonella infection model [69]. For each C. albicans strain, 10 5 cells were injected into 20 Galleria larvae (6 th instar: BioSystems Technology, Exeter, UK). Sterile PBS was injected into control larvae. Survival was monitored for 5 days at 37˚C, represented using Kaplan-Meier curves, and analysed using log rank tests.

Ethics statement
All animal experiments were conducted in compliance with United Kingdom Home Office licenses for research on animals, and were approved by the University of Aberdeen Ethical Review Committee (project license number PPL 70/8583). Animal experiments were minimised, and all animal experimentation was performed using approaches that minimised animal suffering and maximised our concordance with EU Directive 2010/63/EU.
Power analyses based on data generated in previous experiments were applied to estimate the minimum number of animals per group required to achieve statistically robust differences (P <0.05). The power analyses to determine group size for the short term systemic infection model were based on the variation in fungal burdens between animals, whereas those for the long term model were based upon mean survival times. Animals were monitored at least twice daily for signs of distress, which was minimised by expert handling. Euthanasia was performed humanely by cervical dislocation when animals showed signs of progressive illness (e.g. ruffled coat, hunched posture, unwillingness to move and 20% loss of initial body weight). During these studies there were no unexpected deaths. Analgesia and anaesthesia were not required in this study.

Statistical analysis
Statistical analyses were performed in GraphPad Prism 5 and IBM SPSS Statistics (v24.0.0). Two tailed Mann-Whitney U analysis was used to test the statistical difference between two sets of data with a non-parametric distribution. Associations between growth parameters, such as doubling time, lag phase or propidium iodide staining, were determined by one-way and two-way ANOVA and Dunnett post-hoc t-tests. Unstressed samples were used as controls and the values of other samples were compared against these controls. The following p-values were considered: Ã p < 0.05; ÃÃ p <0.01; ÃÃÃ p < 0.001; ÃÃÃÃ p < 0.0001.
Supporting information S1 Fig. Stress resistance of C. albicans tetON-CAT1 and ACT1-CAT1 strains. C. albicans cultures were pre-grown in YPD with 20 μM doxycycline, dilutions spotted onto YPD plates containing H 2 O 2 and/or NaCl stresses at the specified concentrations, and photographed after 24 h growth at 30˚C: CAT1/CAT1, Ca674; CAT1/cat1Δ, Ca1862; cat1Δ/cat1Δ, Ca1864; ACT1 p -CAT1, Ca2031; cat1Δ tetON-CAT1, Ca2038; CAT1 tetON-empty, Ca2084; cat1Δ tetON-empty, Ca2089 (S1 Table). (PDF) S2 Fig. Gating strategy for the analysis of C. albicans Cat1-GFP expressing cells. Exponential populations of C. albicans CAT1-GFP cells (Ca2213: S1 Table) growing in YPD at 30˚C were subjected to fluorescence activated cell sorting. (A) First, singlets were selected and doublets excluded by analysing the FSC signals height versus area. (B) Next, cells of similar size were selected by analysing the FSC versus SSC. (C) This cell population was analysed for their GFP fluorescence intensity (530-540 nm) by plotting GFP against a dump channel (610-620nm). CAT1-GFP cells were compared with control cells with no GFP (Ca674). CAT1-GFP cells with relatively low levels of Cat1-GFP (pink), and cells with relatively high Cat1-GFP levels (cyan) were sorted using the single cell modus of a BD Influx sorter. (D) FACS sorting of C. albicans cells expressing relatively low or high levels of Cat1-GFP in the absence of stress (same figure as Fig 4D).  Table) behaved differently in vivo: isolate 1 displayed decreased colonisation in certain tissues (Fig 5), whereas isolates 4 and 10 did not (see text). Therefore, we tested whether isolates 4 and 10 had lost their phenotype over time. To achieve this we compared the "old" isolates 1, 4 and 10 (Ca2038, Ca2041, Ca2044) with "new" isolates (Ca2040, Ca2043, Ca2046). (Throughout this study, all experiments were performed on strains that had been freshly plated from frozen -80˚C stocks. "Old" isolates were taken routinely from the same -80˚C tubes over a twelve month period. These stocks were not thawed: small volumes were chipped from their frozen surface and plated. Nevertheless, the temperature of these tubes must have increased transiently from -80˚C at regular intervals over this period. In contrast, "new" isolates were from identical -80˚C stocks that were not touched over this period.) (A) As described in Fig 1, catalase activities were measured in C. albicans cells grown in YPD containing 0 or 20 μM doxycycline (-or + Dox, respectively): cat1Δ, Ca2089; wildtype, WT, Ca2084; blue, old tetON-CAT1 isolates, Ca2038, Ca2041, Ca2044; red, new tetON-CAT1 isolates, Ca2040, Ca2043, Ca2046 (S1 Table). Wild-type and cat1Δ cultures were exposed to 0 or 5 mM H 2 O 2 for one hour before analysis. Means and standard deviations from three independent replicate experiments are shown, and the data were analysed using one-way ANOVA with Tukey's post-hoc test: Ã , p 0.05; ÃÃ , p 0.01; ÃÃÃ , p 0.001; ÃÃÃÃ , p 0.0001. (B) The fitness of these old and new tetON-CAT1 strains was compared in vitro by examining their growth (biomass formation; final OD 600 ) on YPD containing or lacking doxycycline (no stress). All of the isolates displayed similar growth in the absence of doxycycline, and the wild-type (CAT1) controls remained unaffected by doxycycline. The presence of doxycycline led to decreased growth for all of the new C. albicans tetON-CAT1 strains and for the old tetON-CAT1-1 strain (Fig S4B). No significant effect of doxycycline on growth and fitness was observed for the old C. albicans tetON-CAT1 isolates 4 and 10. This correlated with a reduction in catalase levels in these isolates over time (Fig S4A). These data indicate that the old isolates 4 and 10 had indeed lost their in vitro fitness defect over time, thereby explaining their lack of phenotype in vivo.  Table). (B) Iron supplementation restores the ability of doxycycline-treated tetON-CAT1 cells to compete with wild type C. albicans cells in mixed cultures. Wild type (Ca2084) and tetON-CAT1-1 (Ca2040) cells were mixed in roughly equal proportions and grown for 12 h in YPD containing 20 μM doxycycline plus 0 or 500 μM FeCl 3 . The proportion of each cell type at the end of these competition experiments (% colonies) was determined by replica plating single colonies onto YPD plates lacking or containing combinatorial stress (1 M NaCl plus 5 mM H 2 O 2 ): wild type cells are sensitive whilst tetON-CAT1 cells resistant to this stress. Means and standard deviations from three replicates are presented: Ã , p 0.05; ÃÃ , p 0.01. (PDF) S1