Phosphoric Metabolites Link Phosphate Import and Polysaccharide Biosynthesis for Candida albicans Cell Wall Maintenance.

Candida species cause hundreds of thousands of invasive infections with high mortality each year. Developing novel antifungal agents is challenging due to the many similarities between fungal and human cells. Maintaining phosphate balance is essential for all organisms but is achieved completely differently by fungi and humans. A protein that imports phosphate into fungal cells, Pho84, is not present in humans and is required for normal cell wall stress resistance and cell wall integrity signaling in C. albicans. Nucleotide sugars, which are phosphate-containing building block molecules for construction of the cell wall, are diminished in cells lacking Pho84. Cell wall-constructing enzymes may be slowed by lack of these building blocks, in addition to being inhibited by drugs. Combined targeting of Pho84 and cell wall-constructing enzymes may provide a strategy for antifungal therapy by which two sequential steps of cell wall maintenance are blocked for greater potency.

fungi and humans. A protein that imports phosphate into fungal cells, Pho84, is not present in humans and is required for normal cell wall stress resistance and cell wall integrity signaling in C. albicans. Nucleotide sugars, which are phosphate-containing building block molecules for construction of the cell wall, are diminished in cells lacking Pho84. Cell wall-constructing enzymes may be slowed by lack of these building blocks, in addition to being inhibited by drugs. Combined targeting of Pho84 and cell wall-constructing enzymes may provide a strategy for antifungal therapy by which two sequential steps of cell wall maintenance are blocked for greater potency.
KEYWORDS Candida albicans, Pho84, antifungal agents, cell wall, chitin synthase, glucan synthase, nucleotide sugar, phosphate metabolism C andida species are the most commonly isolated invasive human fungal pathogens.
Only 3 drug classes are currently available to treat invasive candidiasis, whose attributable mortality is estimated at 19 to 24% (1). Among them, echinocandins, inhibitors of the enzyme beta-1,3-glucan synthase that produces a major cell wall component of Candida species, are now recommended as first-line therapy, since they are candidacidal and have few adverse effects or drug interactions (2,3). Still, outcomes of invasive candidiasis are often poor (2). In fact, early biochemical studies showed that enzymatic activity of beta-1,3-glucan synthase is inhibited by no more than 80% by the echinocandins (4), i.e., echinocandins do not completely inhibit their target (5). Potentiating their effect could be one strategy to improve outcomes in this fearsome infection.
A major barrier to development of new antifungal drugs is the high degree of conservation of many potential drug targets between fungi and humans. We recently found that genetic or pharmacologic interference with the activity of a Candida albicans cell surface phosphate (P i ) transporter, Pho84, which has no human homolog, can indirectly inhibit TOR complex 1 (TORC1) and thereby selectively target fungal proliferation (7). Loss of Pho84 activity also sensitizes C. albicans to oxidative stress and potentiates the activity of two antifungal agents, the polyene amphotericin B and the echinocandin micafungin (7,8).
Micafungin inhibits C. albicans beta-1,3-glucan synthase (9). Questioning how Pho84 activity is related to this enzyme, we found that cells lacking PHO84 poorly tolerated each cell wall stress that we examined. Responsiveness of their cell wall integrity (CWI) pathway signaling through Mkc1 was reduced compared to the congenic wild type. Unlike their oxidative stress hypersensitivity phenotypes (8), homozygous null mutants in PHO84 (pho84 Ϫ/Ϫ ) did not recover cell wall stress resistance by overexpression of the TORC1-activating GTPase, Gtr1. Hence, cell wall stress hypersensitivity of cells lacking Pho84 was mechanistically distinct from their susceptibility to oxidative stress.
Our hypothesis predicts that cells lacking P i or Pho84 activity are also hypersensitive to chemical or genetic perturbation of chitin synthase and beta-1,6-glucan synthase, in addition to beta-1,3-glucan synthase inhibition. We focused on the former two biosynthetic processes, since they may be amenable in the future to pharmacologic inhibition, while excellent beta-1,3-glucan synthase inhibitors are already clinically available or in development (19). Lack of Pho84, P i starvation, genetic depletion of these enzymes, and pharmacologic inhibition of chitin synthase had combinatorial effects on C. albicans growth. Pharmacologic inhibition of Pho84 therefore might potentiate not only echinocandin antifungal activity but also that of other inhibitors of cell wall biosynthetic enzymes.

RESULTS
A Pho84 contribution to cell wall stress resistance was independent of TORC1. C. albicans PHO84 mRNA levels are upregulated during the interaction with host cells as found by others in ex vivo models (20,21) and in vivo during experimental infections (22,23). Cells lacking Pho84 are attenuated in virulence (8). We tested the responses of mutants in PHO84 to stresses they might encounter during infection. Having found that loss of Pho84 potentiates the activity of the echinocandin micafungin (7), we asked whether Pho84 has a role in tolerance of other cell wall stressors which do not act through beta-1,3-glucan synthase inhibition. Congo red is an anionic dye which is thought to disrupt fungal cell wall assembly by binding to the cell wall polysaccharide chitin and disrupting the enzymatic reactions that connect chitin to the glucans, thereby weakening the cell wall (24). Cells lacking Pho84 poorly tolerated chemical cell wall stress induced by Congo red exposure (24) and physical stress induced by heat exposure with osmotic rescue (25,26) (Fig. 1A). These findings indicate that pho84 Ϫ/Ϫ mutants are hypersensitive to cell wall stressors that act by diverse mechanisms.
We previously found an activating activity of Pho84 toward TORC1 to be required for rapamycin tolerance (7). Rapamycin hypersensitivity and Sod3 depletion phenotypes of pho84 Ϫ/Ϫ cells can be suppressed by overexpression of GTR1, which encodes a GTPase component of the TORC1-activating EGO complex (7). GTR1 overexpression had no effect on cell wall stress hypersensitivity of pho84 Ϫ/Ϫ cells (Fig. 1B), suggesting that Pho84 is required for cell wall stress resistance independently of its TORC1-activating role.
Phosphate in Candida albicans Cell Wall Maintenance ® the phosphorylation state of its mitogen-activated protein (MAP) kinase, Mkc1 (28). Mkc1 phosphorylation in response to micafungin was weak in cells lacking Pho84 in P i -poor medium (Fig. 1C), indicating defective activation of CWI signaling. Decreased Mkc1 phosphorylation was less pronounced but still detectable in rich complex medium yeast extract-peptone-dextrose (YPD), which contains 2 mM inorganic phosphate (P i ) as well as accessible organic phosphate compounds (29) (see Fig. S1 in the supplemental material), suggesting that lack of Pho84 impacts CWI signaling even in environments of higher P i abundance. C. albicans TORC1 inhibition is known to upregulate Mkc1 phosphorylation (30), so increased sensitivity of pho84 Ϫ/Ϫ mutants to cell wall stress and their decreased CWI signaling are apparently due to a TORC1independent mechanism.
Cell walls of pho84 ؊/؊ mutants showed decreased alcian blue staining, consistent with decreased cell wall phosphomannan. A major component of C. albicans cell walls is phosphomannan. Phosphodiesters link oligomannosides to glycosylated proteins on the cell surface (31), forming a fibrillar outer layer exterior to the strong chitin and glucan mesh that forms the structural inner layer of the cell wall (32). The phosphomannan component of the C. albicans cell wall confers a negative charge, which can be quantified by binding of the cationic dye alcian blue (33,34). We questioned whether cells confronting a limiting P i supply might prioritize its use for essential metabolic processes over cell wall construction. Using control cells deleted for MNN4, which lack phosphomannan (34), we found decreased alcian blue staining of pho84 Ϫ/Ϫ cell walls during growth at replete (7.3 mM, the concentration of standard synthetic complete medium [SC]) and excess (11 mM) P i concentrations. At moderate P i concentrations (1 mM), the wild type also produced less cell wall phosphomannan, and the difference between pho84 Ϫ/Ϫ and wild-type cells was within the sensitivity range of the assay ( Fig. 2A). In transmission electron microscopic (TEM) images, the outer cell wall layer of pho84 Ϫ/Ϫ as well as pho84 Ϫ/ϩ and pho84 Ϫ/Ϫ/ϩ mutant cells was thinner than that of wild-type cells ( Fig. 2B and C), suggesting that these cells' diminished phosphomannan measurably perturbs the cell wall architecture.
Pho84 was required for production of cell wall polymer precursors. Mannosylation of proteins and addition of phosphomannan occur in the Golgi apparatus where GDP-mannose is the sugar donor (31). The importance of GDP-mannose availability is evinced by the finding that transport of GDP-mannose into the Golgi lumen is the rate-limiting step in cell surface protein mannosylation (31). More generally, nucleotide sugars are precursors of cell wall polymers, e.g., UDP-glucose for beta-1,6-and beta-1,3-glucan (5,16) and UDP-GlcNAc for chitin (35). To examine whether loss of Pho84 affects the availability of biosynthetic intermediates required for cell wall production, we compared metabolomes of wild-type and pho84 Ϫ/Ϫ cells recovering for 4 h in synthetic complete medium (SC) with low (0.22 mM) or excess (11 mM) KH 2 PO 4 from P i starvation, induced by 3 days' incubation in SC medium without P i (SC-P i ) as in the work of Popova et al. (36). This pregrowth period was important in order to unmask P i starvation effects that are otherwise buffered by vacuolar polyphosphate stores in Saccharomyces cerevisiae (37) and are predicted to act similarly in C. albicans (38). Cells grown in this manner, washed in sterile water three times, were extracted three times in 80% methanol for 40 min at Ϫ80°C; supernatants were pooled and dried in a SpeedVac and then stored at Ϫ80°C until analysis.
Hydrophilic interaction liquid chromatography-mass spectrometry (LC-MS/MS) was performed as described in reference 39 to quantitate 258 known metabolites, comparing pho84 Ϫ/Ϫ cells with wild type. We identified significantly altered compounds and pathways using MetaboAnalyst (40). Principal-component analysis showed clustering of genotypes and of ambient P i availability (Fig. 3A). High reproducibility between experiments and significant differences between wild-type and pho84 Ϫ/Ϫ cells were observed by arraying the metabolites measured in three biological replicates (from cells grown on different days) in a heat map (Fig. 3B). The heat map illustrates data values scaled by metabolite abundance in arbitrary units across treatments for each metabolite feature. Similarity measure is based on Euclidean distance and clustered using Ward's linkage.
The genotypes, pho84 Ϫ/Ϫ mutant and wild type, clustered together primarily (Fig. 3B). Low and excess P i concentrations between the two genotypes did not cluster. Closer inspection showed that within some clusters of metabolites, highlighted by rectangles in Fig. 3B, the relative intensities of metabolites were more similar among pho84 Ϫ/Ϫ cells in low P i and wild type in excess P i and vice versa than among similar ambient P i concentrations between the two genotypes. Possibly, some metabolic alterations in cells without Pho84 activity may not be due just to lack of P i but also to aberrant regulatory responses.
Biological process enrichment analysis revealed pyrimidine biosynthesis as among the most significantly affected metabolic pathways during growth of pho84 Ϫ/Ϫ mutant cells at both low and excess ambient P i (Fig. 3C). Purine biosynthesis was also highly significantly altered, as was nucleotide sugar metabolism. We concluded that loss of Pho84 disturbs metabolism of compounds required in cell wall polymer biosynthesis.
Among individual metabolites, purine and pyrimidine nucleotide levels were decreased in pho84 Ϫ/Ϫ cells compared with wild type, while the bases uracil and cytosine and the nucleosides cytidine and guanosine were substantially increased in pho84 Ϫ/Ϫ cells (see Table S1 in the supplemental material). For nucleotide products derived from each nucleobase, we observed accumulation of metabolites before a phosphorylation step and their depletion after this step (Fig. 3D). The most important phosphoric nucleotide precursor, phosphoribosyl-pyrophosphate (PRPP), was sharply decreased (Fig. 3E). These results were obtained using MetaboAnalyst; the column-wise means of all samples from pho84 null mutant cells were divided by the column-wise means of all samples from wild-type cells before column normalization; absolute value changes were compared as fold change. The constellation of metabolic intermediates that we observed suggested that lack of nucleotides was due to a dearth of P i , since interme-diates destined for phosphorylation seemed to have accumulated upstream of the cognate kinase. Degradation products of purines like allantoin were strongly decreased, suggesting that purine salvage was highly upregulated (Table S1).
To test the apparent lack of nucleotides independently, we measured the concentration of the most important nucleotide, ATP, in cells grown to saturation in rich complex medium as in the work of Grahl et al. (41). pho84 Ϫ/Ϫ cells contained substantially less ATP than wild-type or Ϫ/Ϫ/ϩ PHO84 reintegrant cells (Fig. 3F), confirming the result of the LC-MS/MS experiments. Metabolic derangements of pho84 Ϫ/Ϫ cells were extensive and involved multiple further biosynthetic pathways ( Fig. 3C and Table S1). Since ATP participates in a majority of metabolic processes (42), its decreased availability could drive many of the metabolic effects that we observed.
Loss of Pho84 decreased the amount of detectable cell wall chitin. UDP-GlcNAc, the substrate of chitin synthases, was the most significantly decreased cell wall pre- cursor requiring a pyrimidine in pho84 Ϫ/Ϫ cells (Table S1). We hypothesized that if P i to produce specific pyrimidine nucleotide sugars is insufficient, the cell wall polymers produced from these nucleotide sugars will be diminished. To measure cells' chitin content, we modified an assay described in reference 43, using wild-type cells with decreased chitin content for validation. Cells were grown overnight in SC with 0.5 mM P i , unexposed or exposed to increasing concentrations of the chitin synthase inhibitor nikkomycin Z (nikkomycin) (Fig. S2A) (44,45), a competitive inhibitor of C. albicans chitin synthases (44). Fluorescence intensity of calcofluor white-stained cells, recorded by flow cytometry, clearly reflected the decreased chitin content and dose response of nikkomycin-exposed cells (Fig. S2A).
We then measured the chitin content in wild type, pho84 Ϫ/Ϫ mutant, and Ϫ/Ϫ/ϩ PHO84 reintegrant cells recovering from P i starvation, as in the metabolomics experiments, during growth in SC with low (0.22 mM) P i for 4 h. Cells without Pho84 had a significantly lower chitin cell wall content than wild-type and reintegrant cells ( Fig. 4A and Fig. S2B). Chitin quantitation results were highly numerically reproducible among biological replicates performed on different days (Fig. 4A), when measurements were normalized to the mean fluorescence of wild-type cells harvested at the end of P i starvation (time zero). This result suggests a robust metabolic or signaling-based regulatory system that directs biosynthetic fluxes in P i -starved pho84 Ϫ/Ϫ cells away from chitin production and provides a possible causal link to the role of Pho84 in cell wall stress resistance.
Fungal cell wall components are classically analyzed from 2 fractions, alkali insoluble and soluble; the major alkali-insoluble fraction represents a mesh of chitin fibrils covalently linked to glucans (49,50) which provides the structural stability and shape to the cell wall (51). In S. cerevisiae, the covalent bond lending insolubility, in hot NaOH, to this cell wall fraction consists of chitin linked to the nonreducing end of a beta-1,3glucan chain (52). We quantified beta-1,6-glucan in alkali-insoluble cell wall fractions. Alkali-insoluble cell wall fractions from pho84 Ϫ/Ϫ cells contained significantly less beta-1,6-glucan than those from wild-type cells (Fig. 4B). Additionally, we examined alkali-soluble fractions. The alkali-soluble cell wall fraction of C. albicans comprises 5 to 11% of the cell wall mass depending on growth conditions (53) and represents glucans Phosphate in Candida albicans Cell Wall Maintenance ® unlinked to the chitin-glucan mesh that forms the mechanoresistant cell wall core. The alkali-soluble cell wall fraction of pho84 Ϫ/Ϫ mutant cells contained more beta-1,6glucan than that of the wild type (Fig. S3B), suggesting a reduction in covalent linkages among the major cell wall polysaccharides in these cells; the reduction in cell wall chitin content ( Fig. 4A and Fig. S2B) may be responsible for this finding. Overall, major cell wall structural polysaccharides were diminished in cells lacking Pho84 activity, apparently paralleling the availability of their metabolic precursors.
Phosphate deprivation sensitized wild-type cells to pharmacologic inhibition of beta-1,3-glucan and chitin synthesis. If pho84 Ϫ/Ϫ cells are hypersensitive to beta-1,3-glucan synthase inhibition because they lack P i for production of precursors, depriving wild-type cells of P i should have a similar effect. Activity of cell wall polysaccharide-synthetic enzymes decreases when they bind a specific inhibitor (4). By mass action, accumulation of their product should diminish further with declining concentrations of their substrates, i.e., when UDP-glucose and UDP-GlcNAc concentrations drop. Micafungin was used to inhibit beta-1,3-glucan synthase, and nikkomycin was used to inhibit chitin synthase. We had no pharmacological inhibitor of beta-1,6-glucan synthase since the only published such compound (54) is no longer available.
We questioned wild-type cells' responses under conditions that physiologically diminish the role of Pho84, using conditions where its expression in wild-type cells is low. We first established the highest P i concentration at which C. albicans derepresses PHO84 transcription, expecting that, as in S. cerevisiae, PHO84 is repressed in high ambient P i concentrations (37). Using a PHO84 promoter-green fluorescent protein (GFP) fusion, we determined that the PHO84 promoter became derepressed at Յ0.4 mM ambient P i (Fig. S4); hence, we used 0.5 mM as a moderate P i concentration during refeeding of P i -starved cells. Wild-type cells starved for P i in the same way as for the metabolomics experiments, or prefed with excess (12 mM) P i , were reinoculated into moderate (0.5 mM) or excess (12 mM) P i concentrations and exposed to inhibitors of beta-1,3-glucan-and of chitin synthesis. Cells were incubated in 2ϫ SC with 2% glucose in these experiments, in order to optimize nutrients during inhibitor exposure.
Wild type cells prestarved for P i before, and refed moderate P i during, exposure to micafungin were significantly more sensitive than control cells provided with excess P i throughout the experiment (Fig. 5A). However, P i starvation during micafungin exposure did not lead to growth defects of a micafungin-resistant C. albicans bloodstream isolate from a patient treated long-term with this drug (Fig. 5B). This finding indicates that P i starvation did not cause global growth defects in these experiments; it potentiated the effect of specific inhibitors of enzymes whose substrates are linked to P i availability.
Growth defects induced by nikkomycin exposure were enhanced in cells prestarved for P i and refed with moderate P i (Fig. 5C), in wild-type cells as well as in the micafungin-resistant bloodstream isolate (Fig. 5D). Hence, P i starvation sensitized wildtype cells to inhibition of two distinct sugar nucleotide-consuming cell wall biosynthetic processes.
Loss of Pho84 and P i starvation sensitized cells to genetic depletion of chitin and beta-1,6-glucan synthases. If pho84 Ϫ/Ϫ cells' hypersensitivity to cell wall stressors is due to insufficient concentrations of nucleotide sugars, cells with diminished activity of chitin and beta-1,6-glucan synthases should be hypersensitive to loss of Pho84 and to P i starvation. To further test this idea, we perturbed these enzymes genetically.
Among the 4 chitin synthases of C. albicans, the only essential isoenzyme, Chs1, is required for septum production during cell division and contributes to the stability of lateral cell walls (55). We constructed mutants whose only CHS1 allele is controlled by pMAL2 or by tetO, repressible by glucose or doxycycline, respectively, and confirmed that they exhibit previously described phenotypes (13,55) (Fig. S5). The effect of P i availability during CHS1 depletion was examined. CHS1 was depleted from pMAL2 or from tetO after a day of P i starvation or P i excess feeding during which CHS1 expression was induced from these promoters, by incubation in maltose or in the absence of doxycycline. CHS1-depleted cells, incubated in glucose or doxycycline, respectively, had a significant growth defect in a moderate P i concentration (0.5 mM) ( Fig. 6A and B and Fig. S6). The specificity of the P i -dependent growth defect of CHS1-depleted cells was demonstrated by comparatively robust growth of these cells fed excess P i ( Fig. 6A and B).
The role of Pho84 in cells depleted for CHS1 was probed. Loss of PHO84 potentiated the growth defects of cells depleted of CHS1 even in cells fed excess P i (Fig. 6A and B). These experiments suggest that as Chs1 activity became limiting because of a decline of its cognate transcript, P i availability impacted growth significantly. Additionally, a Pho84-specific role seemed to emerge that was independent of ambient P i concentrations.
We constructed strains in which a single allele of the gene encoding the major beta-1,6-glucan synthase, KRE6 (16,18), is controlled by pMAL2. Additionally, the gene encoding the second known beta-1,6-glucan synthase, SKN1, was deleted in the kre6/pMAL2-KRE6 background (18). We observed more severe growth and filamentation phenotypes than did Han et al. (18) during exposure of 2 independent kre6/KRE6 heterozygous deletion mutants to calcofluor white ( Fig. 7A and B), and during depletion of KRE6 from pMAL2 in glucose. To reexamine our findings in light of these discordant results, we constructed strains in which a single KRE6 allele is transcribed from repressible tetO. While neither of these repressible promoters can completely shut off transcriptional activity (56)  Growth of cultures exposed to vehicle or 40 ng/ml micafungin of the wild type (JKC915) or a micafungin-resistant bloodstream isolate (MicaR, JKC2490), pregrown in 2ϫ SC without or with excess P i for 1 day and inoculated into 2ϫ SC with moderate (Modr) or excess (Exs) P i , respectively. (C and D) Growth of cells exposed to vehicle or 8 M nikkomycin of the wild type (JKC915) or a micafungin-resistant bloodstream isolate (JKC2490) treated as in panels A and B. Modr P i , pregrown in 0 P i , inoculated to 0.5 mM P i with vehicle or drug; Exs P i , pregrown with 12 mM P i , inoculated to 12 mM P i with vehicle or drug. Representative of 3 biological replicates; error bars show SD for 3 technical replicates.
Phosphate in Candida albicans Cell Wall Maintenance ® of SKN1 contributed little to growth defects of KRE6-depleted cells under these experimental conditions (Fig. 7C and D).
To examine the effect of P i availability on cells with decreased beta-1,6-glucan synthase activity, cells were P i starved as for the metabolomics experiments but in 2ϫ SC to allow for maximal growth and recovered in moderate (0.5 mM) or excess (12 mM) P i . Control cells were P i loaded in 12 mM P i . Cells depleted for KRE6 after P i starvation and during recovery in moderate P i concentrations had more severe growth defects than cells that were continuously fed excess P i (Fig. 7C and D). This result suggested a need for P i when levels of beta-1,6-glucan synthase became limiting, in order to supply sufficient concentrations of the enzyme's substrate, UDP-glucose.
Loss of Pho84 prevented compensatory chitin deposition in cells depleted for beta-1,6-glucan. Inhibiting beta-1,3-glucan synthesis pharmacologically with echinocandins or depleting beta-1,6-glucan synthases genetically induces compensatory synthesis of chitin by both transcriptional and posttranscriptional mechanisms (57-59). We measured chitin content of cells with and without PHO84, which were depleted for KRE6 with or without intact SKN1 loci (kre6/pMAL2-KRE6 and kre6/pMAL2-KRE6 skn1/ skn1, as well as pho84/pho84 kre6/pMAL2-KRE6 and pho84/pho84 kre6/pMAL2-KRE6 skn1/skn1). Cells were precultured as for metabolomics experiments, except that maltose was provided as the carbon source to permit expression of KRE6, and 2ϫ SC was used to allow for maximal provision of nutrients other than P i . They were then grown for 8 h in 2ϫ SC with 0.22 or 11 mM P i as in metabolomics experiments, using glucose as the carbon source to repress KRE6.
We found a higher chitin content in cells depleted of KRE6, as previously reported (59). This effect was completely or partially abrogated in cells lacking Pho84 (Fig. 8A), depending on the ambient P i concentration. Presence or absence of SKN1 had no effect (Fig. 8A). We concluded that compensatory chitin synthesis in cells whose beta-1,6glucan biosynthesis was diminished required the availability of sufficient P i , as well as an activity of Pho84.

Loss of Pho84 and chitin synthase inhibition potentiated growth defects during beta-1,6-glucan synthase depletion.
When beta-1,3-glucan synthase is inhibited by an echinocandin, upregulation of chitin synthesis can compensate for loss of cell wall stability and inhibition of growth (57). We examined the effects of blocking compensatory mechanisms for loss of beta-1,6-glucan. Beta-1,6-glucan synthase was depleted by repressing KRE6 from pMAL2, and Pho84 activity was eliminated genetically, while chitin synthase was inhibited with low concentrations of nikkomycin. Loss of Pho84 activity and chitin synthase inhibition each further reduced growth of cells lacking beta-1,6-glucan synthase (Fig. 8B). We concluded that the inhibitory effects of beta-1,6-glucan synthase depletion are potentiated by inhibition of chitin synthase and loss of Pho84 activity.

DISCUSSION
Cells lacking Pho84 are hypersensitive to cell wall stress (7) (Fig.1A). Our mechanistic analysis of this phenotype indicated that, unlike their oxidative stress hypersensitivity, were pregrown in 2ϫ SC-2% maltose without or with excess (12 mM) P i for 3 days, and dilutions were spotted on 2ϫ SC-2% glucose containing moderate (0.5 mM) or excess P i . (D) Cells pregrown without or with excess P i as in panel C were inoculated into 2ϫ SC-2% glucose containing moderate or excess P i , respectively. Representative of 3 biological replicates; error bars show SD for 3 technical replicates. ϩ/ϩ, JKC915; pho84 Ϫ/Ϫ , JKC1450; kre6/pMAL2-KRE6, Phosphate in Candida albicans Cell Wall Maintenance ® it was not directly related to these cells' diminished TORC1 signaling (Fig. 1B). While oxidative stress signaling is upregulated in pho84 Ϫ/Ϫ cells (8), their cell wall integrity signaling was abnormally weak as measured by the phosphorylation state of the CWI MAP kinase Mkc1 (Fig. 1C). Dampened CWI signaling in pho84 Ϫ/Ϫ cells cannot be explained by their decreased TORC1 signaling activity, because TORC1 inhibition induces signaling through Mkc1 (30). Decreased alcian blue staining of pho84 Ϫ/Ϫ mutants suggested that the phosphomannoprotein content of their cell walls was decreased ( Fig. 2A). Significant thinning of their outer phosphomannan cell wall layer was found by measurements of TEM images ( Fig. 2B and C). This thinning was far less striking than that seen in TEM micrographs of mutants in enzymes that produce this layer, e.g., in the Mnn2 family of mannosyltransferases (60). Nevertheless, the difference in outer layer thickness between wild-type and pho84 Ϫ/Ϫ cells was highly significant (Fig. 2C). The phenotype was similar in cells lacking one or both copies of PHO84. We noted striking haploinsufficiency of pho84/PHO84 cells (and of pho84/pho84::PHO84 cells) for outer cell wall layer thickness; haploinsufficiency is known to affect both structural and regulatory genes in C. albicans (61,62). We concluded that lack of Pho84 can disturb the normal cell wall architecture.
This finding prompted the idea that cells lacking Pho84 are defective in synthesizing cell wall components that require phosphorylated precursors. Structural cell wall polysaccharides of C. albicans, beta-1,6-glucan, beta-1,3-glucan, and chitin in order of their abundance (16), are synthesized from monosaccharide precursors activated with UTP to generate UDP-containing nucleotide sugars; UTP biosynthesis requires availability of the P i donor ATP. Vacuolar polyphosphate stores buffer decreased extracellular P i availability in S. cerevisiae (37); vacuolar polyphosphate storage pools are also present in C. albicans (38). Hence, we applied the protocol of Popova et al. (36) to neutralize intracellular P i stores before an incubation period in low, moderate, or excess After 3 days' passages in 2ϫ SC-2% maltose without or with excess P i , cells were inoculated into 2ϫ SC-2% glucose with moderate or excess P i , respectively, with vehicle (Veh) or 0.25 M nikkomycin (Nikko). Modr P i , pregrown without P i , inoculated to moderate (0.5 mM) P i ; Exs P i , pregrown with excess (12 mM) P i , inoculated to excess P i . Representative of 3 biological replicates; error bars show SD for 3 technical replicates. ϩ/ϩ, Metabolomics experiments showed derangements of biosynthetic pathways that require phosphorylation steps in pho84 Ϫ/Ϫ cells (Fig. 3). Pyrimidine biosynthesis was highly significantly altered. The sugar nucleotides that act as precursors for cell wall polysaccharide biosynthesis, UDP-glucose and UDP-GlcNAc, were decreased 2.6-and 2.9-fold, respectively, in pho84 Ϫ/Ϫ cells in our experiments (Table S1). Our findings agree with those of Boer et al., who found decreased levels of nucleotides and of the nucleotide sugar UDP-glucose in P i -limited S. cerevisiae cells grown in continuous culture in a chemostat (63).
We considered whether accumulation of toxic metabolites might account for cell wall biosynthesis defects. However, toxic metabolites that accumulate, e.g., in S. cerevisiae models of galactosemia (64,65) and fructose intolerance (65), are sugar phosphates, i.e., their biosynthesis requires a phosphorylation step whose substrate was scarce in cells lacking Pho84. Accordingly, we did not identify potentially toxic metabolites among the significantly dysregulated metabolites in these cells (Table S1).
Perturbation of multiple other metabolic processes was observed. This result is consistent with depletion of ATP, a major P i donor and energy currency of the cell, which we confirmed in independent assays for pho84 Ϫ/Ϫ cells (Fig. 3F). Glycolysis and galactose and pentose processing were disrupted (Fig. 3C), consistent with a requirement for phosphorylation in metabolism of these sugars; glycolysis alone consumes 2 P i and 2 ADP molecules per molecule of glucose (66). Possibly, defects in producing sugar precursors of nucleotide sugars may contribute to the disruption of nucleotide biosynthesis that we observed in cells lacking Pho84. Insufficient ATP to produce nucleotide sugars may also be ultimately responsible for cell wall phenotypes of cells perturbed in mitochondrial function, as reported, e.g., in references 67 to 72. However, ATP deprivation is expected to impact pho84 Ϫ/Ϫ cells' fitness in multiple processes beyond diminished cell wall precursor availability; its effect in C. albicans' different natural niches and life cycle stages remains to be explored in future experiments.
The metabolomics findings suggested a simple model, by which lack of monomeric precursors for cell wall polysaccharides deprives the cognate enzymes of their substrates, slowing the reaction velocity and leading to decreased structural cell wall polysaccharides and hence decreased cell wall stability (Fig. 9). Diminished cell wall carbohydrate content of P i -starved chemostat-grown S. cerevisiae cells was described 4 decades ago, though a mechanism was not proposed (73). Our model predicts that cells lacking Pho84 would be intolerant not just of beta-1,3-glucan synthase inhibition by micafungin (7) but also of genetic or pharmacologic perturbation of beta-1,6-glucan and chitin synthases. The chitin and beta-1,6-glucan contents of cells lacking Pho84 were sharply decreased (Fig. 4). Similarly, wild-type cells starved for P i were hypersensitive to inhibitors of chitin and beta-1,3-glucan synthesis, nikkomycin and micafungin (Fig. 5), consistent with combinatorial effects between depletion of the enzyme substrate and direct enzyme inhibition.
Growth defects of strains that we constructed in which expression of the genes encoding the major beta-1,6-glucan synthase, KRE6, and the single essential chitin synthase, CHS1, is repressible from pMAL2 ( Fig. 6 and 7) or tetO were in agreement with the results of Munro et al. (55) but differed from the findings of Han et al. (18,59). In some kre6/pMAL2-KRE6 strains, we also deleted the minor beta-1,6-glucan synthaseencoding gene, SKN1, and observed no further effect on the phenotype under the analyzed conditions. While transcription from the repressible promoters we used cannot be completely abrogated (56), residual transcription is not expected to sharpen a loss-of-function-associated growth defect. We speculate that the stronger defects of our kre6 conditional mutants, compared to the homozygous deletion mutant of Han et al., may be attributable to residual Kre6 activity in our mutants, which lowered the likelihood of suppressor mutation emergence.
Loss of Pho84 exacerbated the growth defects exhibited by CHS1-and KRE6depleted strains, especially when P i was not in excess in the medium (Fig. 6 and 7). Conversely, maximally loading cells with P i by prolonged growth in excess P i during depletion of the enzyme in question significantly rescued these growth defects ( Fig. 6  and 7), supporting the idea that lack of P i is responsible for synthetic phenotypes of pho84 with kre6 or chs1 mutations.
In agreement with the work of Han et al. (59), we observed a compensatory increase of cell wall chitin (Fig. 8A) in cells depleted of KRE6. Han et al. showed that this response depended on intact Mkc1 signaling (59). Absence of Pho84 in KRE6-depleted cells abrogated this compensatory response (Fig. 8). Dampened Mkc1 signaling in cells without Pho84 activity ( Fig. 1C and Fig. S1) may be one reason for the pho84 Ϫ/Ϫ mutants' inability to appropriately upregulate chitin synthase transcription (58,74). Another reason could be that the concentration of the chitin synthase substrate UDP-GlcNAc was insufficient. The two mechanisms could act together to diminish compensatory chitin synthesis in cells lacking Kre6 as well as Pho84.
Unlike Han et al. (59), we found that very low nikkomycin concentrations further inhibit the growth of KRE6-depleted cells, as compensatory chitin synthesis is inhibited (Fig. 8B). This discrepancy could be due to the medium used for assaying the nikkomycin effect: di-and tripeptides, present in YPD but not in the SC medium we used, compete with nikkomycin for uptake into C. albicans cells (75,76). If novel antifungals could be combined to simultaneously inhibit glucan and chitin synthesis, a potent antifungal effect as well as low toxicity in humans, who lack both targets, could be expected.
Induction of high-affinity P i transporter-encoding genes PHO84 and PHO89 in ex vivo and in vivo models of infection (20)(21)(22)(23) shows that during infection, the fungus is FIG 9 Model: role of Pho84 in cell wall polysaccharide biosynthesis. P i is imported across the cytoplasmic membrane by Pho84, where it is available to metabolic processes that produce the nucleotide sugars UDP-glucose and UDP-GlcNAc. These nucleotide sugars are the substrates of glucan and chitin synthases that produce major polysaccharides of the cell wall, thereby providing its structural stability. challenged with acquiring P i , possibly due to alkaline environments in the host (77). While our experimental conditions of prolonged P i excess are apparently not typical for niches inhabited by C. albicans, our results showed that it is limited P i availability that renders depletion of these cell wall biosynthetic enzymes inhibitory to C. albicans growth. They further indicated that during perturbation of biosynthesis of a single cell wall component, P i availability is limiting for production of compensatory cell wall components like chitin.
Investigators examining C. albicans isolates from stool of intensive care unit patients as well as standard laboratory strains hypothesized that phosphate starvation increases hyphal growth and virulence of C. albicans (78). Mice that had undergone partial hepatectomy and were drinking tap water, versus a 25 mM phosphate solution, were more susceptible to cecal injection of a C. albicans suspension in water than in 25 mM phosphate, respectively, and had more C. albicans biofilm on their intestinal mucosa (78). The authors concluded that phosphate starvation had made the C. albicans cells more virulent. Mutants in the phosphate starvation response transcriptional regulator Pho4 were considered to produce more hyphae on low-phosphate than on highphosphate medium and to be more virulent in a Caenorhabditis elegans model of infection (78). These findings contrast with those of Ikeh et al., who in multiple ex vivo and in vivo infection models found null mutants in PHO4 to be attenuated in virulence (38). Since cells without Pho4 have low levels of Pho84 (38), our results align more closely with those of Ikeh et al. (38), as we found attenuated virulence and defective hyphal growth in cells lacking Pho84 (8). The cell wall integrity defects of pho84 null mutant cells characterized here may also contribute to their virulence attenuation. Differences in experimental conditions may be responsible for the discrepancies between the results reported by these investigators (78) and await analysis in further experiments.
How limited P i supplies and essential intermediate metabolites requiring P i like ATP and phosphoribosyl-pyrophosphate (PRPP) are allocated to different biosynthetic activities of the cell is not known. Decreased phosphomannan, decreased chitin, and decreased beta-1,6-glucan contents suggest that cells lacking P i prioritize use of this essential element for other metabolic processes. P i starvation often limits plant growth: P i -starved oats replace up to 70% of their plasma membrane phospholipids with the glycolipid digalactosyldiacylglycerol (79), a process that is reversible upon P i refeeding (80). Fungi including the human pathogen Cryptococcus neoformans also replace cytoplasmic membrane phospholipids with nonphosphoric lipids during P i starvation (81,82). That the cell envelope-the plasma membrane and cell wall-apparently can function while forgoing a share of P i , while DNA polymerase and ribosomes are absolutely dependent on their P i allotment, suggests a regulatory mechanism to assign the available P i to each biosynthetic process. Identification of and interference with this mechanism might lead to a way to disrupt processes required for growth and proliferation of the fungal cell.
Echinocandins and nikkomycin exemplify good tolerability of antifungals whose targets are not conserved in humans (3,83). Pho84 and beta-1,6-glucan synthases also have no human orthologs. Development and combination of specific small-molecule inhibitors of these targets should potentiate their effects and permit more effective clearance of invasive candidiasis.

MATERIALS AND METHODS
Detailed descriptions of methods are provided in Text S1 in the supplemental material. Strains and culture conditions. C. albicans strains used are shown in Table S2A. Strains were constructed as described in reference 56, using plasmids shown in Table S2B and oligonucleotides shown  in Table S2C, using sequences obtained from the Candida Genome Database (84). To minimize phenotypic artifacts originating from genomic events unrelated to the targeted introduced mutations, all genotypes examined were constructed from at least 2 independently engineered heterozygous strains. Experiments with defined ambient P i concentrations were performed in media based on yeast nitrogen base (YNB) 0 P i (ForMedium Ltd., Norfolk, United Kingdom) with added KH 2 PO 4 to stated concentrations. P i starvation was induced as described in reference 36. Other media were used as previously indicated (56).
Phosphate in Candida albicans Cell Wall Maintenance ® Western blots. Cell lysis and Western blotting were performed as described in reference 30. Antibodies used are listed in Table S2D. At least three biological replicates were obtained.
Alcian blue staining assay. The standard curve and alcian blue binding were determined as in the work of Hobson et al. (34).
Transmission electron microscopy. Each strain was inoculated in standard SC to an optical density at 600 nm (OD 600 ) of 0.1 and grown for 15 h. Cells were then prepared for TEM and analyzed as described in reference 60 with minor modifications. Intracellular ATP measurement. Cells cultured overnight in YPD medium with a starting OD 600 of 0.2 were washed with sterile water and lysed. Intracellular ATP levels were measured using the CellTiter-Glo luminescent cell viability assay (Promega; catalog no. G7570), and the results were normalized to the protein concentration. The standard curve was prepared using ATP disodium salt hydrate (Sigma; catalog no. A6419).
Metabolomics. Metabolites were measured as described in reference 39. Hyphal morphogenesis assay. Cells were resuspended in 0.9% NaCl to an OD 600 of 0.1. Variations between single colonies as well as colony density effects were minimized by spotting 3-l cell suspensions at 6 equidistant points, using a template, around the perimeter of an agar medium plate as described in reference 56. Spider medium and RPMI 1640 with glutamine, without sodium bicarbonate (Gibco 31800-022), were used.
Chitin measurement. Cells were grown for 8 h in SC low (0.22 mM) or excess (11 mM) P i after pregrowth of 3 days' passages in 2ϫ SC-P i 2% maltose. At the end of pregrowth (time zero) and at 8 h, cells were formaldehyde fixed, washed in phosphate-buffered saline (PBS) before staining with calcofluor white, and then washed extensively in 0.9% NaCl. After sonication, fluorescence intensities of cells were measured by flow cytometry of Ն10 5 events.
PHO84 promoter induction analysis. Cells of genotype PHO84/pPHO84-GFP-NAT1-PHO84 were grown in YPD with additional 10 mM P i for 16 h and washed three times with 0.9% NaCl, and OD 600 was adjusted to 0.01 in SC with increasing concentrations of KH 2 PO 4 . OD 600 and GFP signal were recorded every 30 min.
Beta-1,6-glucan measurement. Cell wall glucans were extracted, adapting the method of Gilbert et al. (85), by alkali extracting crude cell lysate with 0.75 N NaOH at 75°C for an hour. Supernatants containing alkali-soluble glucans were stored at Ϫ80°C until analysis by ELISA. Insoluble pellets were digested with chitinase and Zymolyase in 100 mM K 2 HPO 4 -KH 2 PO 4 buffer (pH 6.0) for 72 h at 37°C followed by 1 h at 45°C. Supernatants of these digests containing alkali-insoluble glucans were stored at Ϫ80°C until analysis by ELISA. A competition ELISA using an anti-beta-1,6-glucan antibody (48) was performed with modifications as described in reference 48.
Statistical analysis. Statistical analysis was performed by unpaired Student's t test in Prism 7 GraphPad (GraphPad Software, Inc., CA, USA). For metabolomics, MetaboAnalyst (40) was used for analysis including principal-component analysis, heat maps, and pathway analysis.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, PDF file, 0.1 MB.