Regulation of heme utilization and homeostasis in Candida albicans

Heme (iron-protoporphyrin IX) is an essential but potentially toxic cellular cofactor. While most organisms are heme prototrophs, many microorganisms can utilize environmental heme as iron source. The pathogenic yeast Candida albicans can utilize host heme in the iron-poor host environment, using an extracellular cascade of soluble and anchored hemophores, and plasma membrane ferric reductase-like proteins. To gain additional insight into the C. albicans heme uptake pathway, we performed an unbiased genetic selection for mutants resistant to the toxic heme analog Ga3+-protoporphyrin IX at neutral pH, and a secondary screen for inability to utilize heme as iron source. Among the mutants isolated were the genes of the pH-responsive RIM pathway, and a zinc finger transcription factor related to S. cerevisiae HAP1. In the presence of hemin in the medium, C. albicans HAP1 is induced, the Hap1 protein is stabilized and Hap1-GFP localizes to the nucleus. In the hap1 mutant, cytoplasmic heme levels are elevated, while influx of extracellular heme is lower. Gene expression analysis indicated that in the presence of extracellular hemin, Hap1 activates the heme oxygenase HMX1, which breaks down excess cytoplasmic heme, while at the same time it also activates all the known heme uptake genes. These results indicate that Hap1 is a heme-responsive transcription factor that plays a role both in cytoplasmic heme homeostasis and in utilization of extracellular heme. The induction of heme uptake genes by C. albicans Hap1 under iron satiety indicates that preferential utilization of host heme can be a dietary strategy in a heme prototroph.

Introduction forward genetics. We took advantage of this system to identify new components and regulators of the heme utilization pathway in C. albicans. One of the genes identified, which we characterize here, encodes a zinc finger transcription factor that responds to extracellular heme and regulates both cytoplasmic heme homeostasis and extracellular heme uptake.

Identification of mutants defective in hemoglobin utilization
Identification of mutants unable to utilize hemoglobin was done in two stages. In the first stage, we enriched a pool of haploid cells mutagenized with the Ac/Ds transposon from maize [24] for Ga 3+ -protoporphyrin IX (GaPPIX)-resistant clones. GaPPIX is a toxic heme analog that inhibits the growth of bacteria and yeasts [27,28]; in C. albicans, GaPPIX toxicity depends on a ferric reductase-like protein involved in heme uptake, Frp2 [22]. The mutagenized pool was serially diluted 1:1000 and outgrown five times in YPD medium at pH 7.5, supplemented with 1 mM of the iron chelator ferrozine to elicit expression of heme uptake genes, and containing 150 μM GaPPIX. In a second stage, aliquots from the GaPPIX enrichment pools were spread on YPD plates and then replica-plated on YPD plates supplemented with the iron chelator BPS (bathophenanthroline disulfonic acid) and hemoglobin, a medium that compels the cells to use hemoglobin as iron source. Mutants unable to grow on these plates, i.e. unable to use hemoglobin as the sole iron source, were isolated and characterized for their transposon insertion site. Since we found that the mutant pools after enrichment round 3 all contained insertions in FRP2 only, we screened for hemoglobin utilization mutants only after the first two rounds of enrichment. Table 1 shows the identification of the transposon insertions that were associated with reduced growth on hemoglobin. The vast majority (62/73) fell in eight genes that constitute the RIM pathway, a signal transduction pathway that is activated in neutral or alkaline pH [29] and that was shown to activate, among others, FRP1 [30,31], which is essential for heme-iron acquisition [22]. Three other genes, VPS23, VPS25, VPS36, encode components of the ESCRT pathway, which was shown to be involved in hemoglobin-iron acquisition [17]. Seven other genes occurred once, including FRP2. FRP2 is the gene most strongly selected for by GaPPIX, however FRP2 mutants are generally not defective in heme uptake in unbuffered YPD medium [22]. A single insertion, near the 3' end of the coding region, was an exception, for unknown reasons. One more gene, orf19.4145 / C5_01500C / ZCF20, a predicted zinc finger transcription factor, was identified by two transposon insertions, and is characterized below as a heme-regulated activator of genes for heme uptake and homeostasis. The sequence most similar to Zcf20 in the S. cerevisiae genome is Hap1 (Heme Activator Protein 1), a heme-regulated transcription factor [32][33][34]. However, the similarity between the protein sequences is low, and is mainly concentrated in the DNA binding domain of Hap1 (S1 Fig). Furthermore, a search for Hap1-related sequences in C. albicans and related species of the CUG-Ser1 clade, showed that the Mrr1 transcription factor and its relatives are equally or more closely related than Zcf20 and its relatives to S. cerevisiae Hap1 (S2 Fig). However, closer examination of the sequences revealed that Zcf20 and its relatives contain CP motifs embedded in the consensus K/RCPV/I, which were characterized as heme regulatory motifs (HRM) in Hap1 [35] (S1 and S3 Figs). While Hap1, Zcf20 and all their closest homologs contain multiple repeats of this motif (S3 Fig), the Mrr1-related sequences contain none. Based on this, and based on the analyses shown below, we suggest that ZCF20 is the likely ortholog of S. cerevisiae HAP1, and we therefore propose to rename it C. albicans HAP1.

Hap1 promotes sensitivity to heme analogs and hemoglobin-iron utilization
To confirm the hap1 phenotypes in a standard diploid C. albicans strain, we first took advantage of an existing deletion of HAP1/ZCF20 within a library of transcription factor mutants [36], and found that the hap1 -/strains are defective in growth at low hemoglobin concentrations compared to the wild-type (S4 Fig). For further experiments, we constructed a new set of HAP1 deletion strains and reintegrants in a standard diploid strain background. We tested sensitivity of the mutants to GaPPIX in several ways. First, we replicated the initial screen and grew the wild-type and two independent mutants in shaking cultures with 150 μM GaPPIX. Under these conditions, the hap1 -/deletion strains grew faster than the wild-type (S5 Fig). We then measured the minimal inhibitory concentration (MIC) of GaPPIX as well as of another toxic heme homolog, Co 3+ -protoporphyrin IX (CoPPIX), in 96 well plate format. MIC for GaPPIX was increased several-fold in the hap1 -/mutant strains (Fig 1A), and the hap1 -/mutant appeared to be completely resistant to up to 0.2 mM CoPPIX ( Fig 1B). The hap1 -/mutant was similarly much more resistant to toxic heme analogs in Petri plates (Fig 1C).
We next tested heme-iron utilization of the hap1 -/mutant vs. wild-type or reintegrant strains. The mutant exhibited reduced growth on plates containing 1 μM hemoglobin as the sole iron source (Fig 2A). In liquid medium, growth of the mutant strains was reduced at low concentrations but restored at higher concentrations of either hemoglobin or hemin as iron sources (Fig 2B and 2C).

Hap1 drives hemin-responsive gene expression
To identify potential Hap1 target genes in the C. albicans genome, we scanned the intergenic regions of the C. albicans genome for the proposed consensus binding site of S. cerevisiae , two independent hap1 -/deletion strains (KC1408) and the reintegrated strains (KC1407). The indicated strains were inoculated in triplicate with the indicated GaPPIX concentrations in 96 well plates, in YPD + 1 mM ferrozine to induce the heme-uptake genes, and incubated for 2 days at 30˚C. The graph indicates the average density for each triplicate, and the error bars indicate the standard deviations. (B) CoPPIX MIC assay of the same strains as in A. (C) Drop dilution assay of the same strains, inoculated on YPD medium supplemented with 1 mM ferrozine and 0.1 mM CoPPIX, vs. a YPD control plate. The Petri plates were incubated at 30˚C for two days.
https://doi.org/10.1371/journal.pgen.1010390.g001 Hap1 (CGGNNNTANCGG [35,37]), using the Patmatch function of the Candida Genome Database site. We found 13 sites, including one near FRP2 (pos. -324) and one near HAP1 itself (pos. -878). The heme oxygenase HMX1, an additional candidate Hap1 target gene based on its function, has a near-consensus site at -206. Based on the presence of the S. cerevisiae Hap1 consensus site, we surmised that these genes might be activated by Hap1 in the presence of hemin.

PLOS GENETICS
To test Hap1-dependent expression of these genes in the presence of hemin, a wild-type strain and two hap1 -/mutant strains were exposed to 50 μM hemin and samples were taken for qRT-PCR at 0, 1 h and 2 h after hemin addition. The results shown in Fig 3 indicate that in the wild-type strain, all three genes (FRP2, HMX1 and HAP1) are induced in the presence of hemin. Furthermore, HMX1 expression is largely dependent upon Hap1even in the absence of hemin. In contrast, FRP2 is expressed in the absence of Hap1, but at lower levels than in the wild-type strain. As expected, no HAP1expression is detected in the hap1 -/mutant. However, HAP1is itself induced in the presence of hemin in the medium.

Hap1 regulates heme homeostasis and heme influx into the cell
The effect of Hap1 on expression of FRP2 suggests that it is involved in heme uptake, whereas its effect on the heme oxygenase HMX1 suggests that it is involved in cytoplasmic heme homeostasis. To test the influence of Hap1 on cytoplasmic heme levels in the absence and presence of added heme in the medium, we used a recently developed cytoplasmic fluorescent heme sensor (HS1). This sensor is comprised of a green fluorescent protein (GFP) domain into which a heme-binding cytochrome domain is embedded, as well as a separate red fluorescent protein (mKATE2) domain. Binding of heme to the cytochrome domain causes quenching of the GFP fluorescence signal but not of the mKATE2 signal; hence, the GFP/mKATE2 fluorescence ratio gives an estimate of cytoplasmic "free" heme levels (or more accurately, "labile" or "exchangeable" heme levels [38,39]), with lower ratios corresponding to higher heme levels [14,40]. To extend the dynamic range of detection, a second sensor, with lower affinity due to a mutation (M7A) in the cytochrome domain, is also used.
We first compared wild-type HAP1 cells to two independent hap1 -/mutant strains, each transformed with the two heme sensor plasmids. As expected, the high-affinity HS1 sensor in unperturbed cells gives a lower fluorescence ratio, corresponding to a higher heme occupancy, than the medium-affinity M7A sensor ( Fig 4A). Furthermore, in the absence of added hemin to the medium, the hap1 -/strains exhibited 25% lower fluorescence ratios with the high-affinity HS1 sensor and 15% lower fluorescence ratios with the medium-affinity M7A sensor, small but statistically significant differences that indicate the presence of higher heme concentrations in the cytoplasm of the hap1 -/cells ( Fig 4A). When heme was added to the medium at increasing concentrations, the fluorescence ratios decreased in all cells, indicating influx of heme into the cytoplasm. At all hemin concentrations, the hap1 -/cells exhibited lower fluorescence ratios, i.e. higher cytoplasmic heme concentrations, compared to the wild-type: in the presence of 2.5 μM hemin in the medium, the high and medium affinity sensors gave 40% and 20% lower signals, respectively, in the hap1 -/mutant strains vs. the wild-type strain. In 10 μM and
We hypothesized that the higher heme levels detected in the hap1 -/strains might be because Hap1 affects the expression of HMX1, which encodes a cytoplasmic heme-degrading enzyme (Fig 3). To test this hypothesis, we deleted HMX1 in both the HAP1 wild-type and mutant strains, and compared heme sensor activity in the absence and presence of hemin in the medium. In the absence of hemin, the HS1 sensor gave identical signals in the HAP1 wildtype and mutant strains, suggesting that the effect of HAP1 on "baseline" cytoplasmic heme levels is due to its effect on HMX1 expression (Fig 4B). We also noted that the HS1 signals were lower in the hmx1 -/vs. the HMX1 strains (compare panels A and B in Fig 4), corresponding to higher baseline cytoplasmic heme levels in the hmx1 -/cells, as had been described before [14].
We next tested the effect of addition of 5-50 μM of hemin to HAP1 and hap1 -/strains in the hmx1 -/background. The high-affinity HS1 sensor became saturated in both the HAP1 and hap1 -/strains even at the lowest hemin concentration (Fig 4B). The medium-affinity M7A sensor also became saturated at all hemin concentrations in the HAP1 strain, indicating that the combination of heme influx and lack of degradation by Hmx1 lead to high cytoplasmic heme concentrations. However the hap1 -/strain gave a different picture: there, decreasing signals were detected with increasing external hemin, but except for the highest hemin concentration (50 μM), the signals were significantly higher in the hap1 -/strain than in the HAP1 strain, indicating reduced influx of heme into the cell ( Fig 4B).
Taken together, the data indicate that Hap1 affects cytoplasmic heme homeostasis via its effect on expression of the heme oxygenase HMX1. Once HMX1 is removed, a second effect of Hap1 is revealed, on heme influx: in the hap1 -/mutant, less heme appears in the cytoplasm of cells exposed to hemin in the medium than in the HAP1 wild-type.

Hap1 is nuclear localized and stabilized in the presence of hemin
To investigate the regulation of Hap1 by heme, we fluorescently tagged the protein at its C-terminus by integration of GFP into one of the two HAP1 alleles. Cells expressing Hap1-GFP were grown in iron-limiting medium (YPD + 1 mM ferrozine), which induces HAP1 /orf19.4145 [41], as well as in the presence of hemin, which was shown here to induce HAP1 (Fig 3), and compared to growth in regular YPD medium. Both iron limitation and the presence of hemin in the medium did induce a fluorescence signal that appeared to be nuclear, whereas the signal was barely visible in regular medium ( Fig 5A). Comparison with a nuclear stain confirmed that the GFP signal induced by hemin is indeed concentrated in the nucleus ( Fig 5B). Since HAP1 affects sensitivity to the toxic heme analogs GaPPIX and CoPPIX, we also tested induction of the nuclear Hap1-GFP signal with these and other heme analogs. As shown in Fig 5C, at an identical concentration of 50 μM, hemin gave the strongest induction, but other heme analogs also induced Hap1-GFP, with Zn 2+ -protoporphyrin IX (ZnPPIX) and CoPPIX giving stronger signals than GaPPIX, while Mn 3+ -protoporphyrin IX (MnPPIX) did not appear to induce Hap1-GFP at this concentration.
The induction of the Hap1-GFP signal could be due to induction of HAP1 transcription, but we surmised that the protein could also be directly affected by the presence of heme. We therefore measured the stability of the protein by placing an epitope-tagged version of HAP1 under the glucose-repressible MAL2 promoter and measuring protein levels after addition of glucose. As shown in Fig 5D, while in regular medium, after an initial lag, Hap1decayed with a half-life of 20-30 min, in the presence of hemin in the medium, it was completely stabilized. Thus, we conclude that stabilization of the Hap1 protein contributes to the increase in Hap1 levels in the presence of hemin in the medium. , or a control strain (C = SN148) were grown overnight in YEP+ 2% raffinose at 30˚C. In the morning, cells were diluted to OD 600 = 0.5 in YEP+2% maltose, with or without 20 μM hemin, and incubated for 4 hours at 30˚C. Next, 2% glucose was added, and the Hap1-6xMyc protein levels were followed by Western blotting. Time 0' is immediately after glucose addition. Hap1-6xMyc was detected with the anti-Myc antibody 9E10, and the same membrane was reacted with an anti-tubulin antibody. Protein signals were visualized by chemoluminescence and quantitated. The left panel shows the membrane, and the right panel shows the ratio of Myc signal to tubulin signal, normalized to the signal at time 0' for each culture (= 100%). The graph shows the average of two separate experiments. The error bars indicate standard deviations. https://doi.org/10.1371/journal.pgen.1010390.g005

Role of Hap1 in global gene expression in the presence of hemin
To identify the genes activated by Hap1 in the presence of hemin in the medium, hap1 -/cells transformed either with a vector plasmid or with a plasmid harboring the wild-type HAP1 gene were exposed to 50 μM hemin in the medium for 30 min, and their global transcriptome was analyzed by RNA-SEQ (S1 Table). Expression was compared between HAP1 and hap1 -/cells with or without added hemin, and between hemin-supplemented medium and control medium in both types of cells (see S2 and S3 Tables for lists of genes significantly induced or repressed, respectively).
To obtain an overview of the types of genes affected by hemin or by HAP1, we performed GO enrichment analysis by biological process on all genes induced or repressed 2x or more in HAP1 reintegrant vs. hap1 -/mutant cells with or without added hemin, and in hemin-supplemented vs. regular YPD medium in both types of cells (S4 and S5 Tables). Fig 6A lists the top GO terms of the Biological Process hierarchy in each comparison. Addition of hemin in the HAP1 strain caused an increase, among induced genes, in "oxidant detoxification", consistent with the notion that free heme can cause oxidative damage to membranes and proteins [3], as well as "response to host immune response", whereas hemin addition caused an increase in "carbohydrate transport" both in the presence and absence of Hap1. Addition of hemin also caused repression of ribosome biogenesis., both in the presence and absence of Hap1.
When comparing HAP1 and hap1 -/cells, "cellular iron homeostasis" stands out among induced genes, probably reflecting the heme-uptake genes induced by Hap1, as detailed below. In addition, in HAP1 vs. hap1 -/cells in regular medium, rRNA biogenesis also appears among the "induced" genes-i.e., repressed in hap1 -/--, whereas categories representing ATP generation by glycolysis, carbohydrate transport and responses to toxic chemicals appear among the repressed genes-i.e., induced in hap1 -/-. In the presence of added hemin in the medium, in contrast, most of these GO categories are not overrepresented anymore in HAP1 vs. hap1 -/cells.
The overlap between the four subsets of genes induced at least 2x in all four pairwise comparisons was plotted in Fig 6B. Only seven genes were induced by hemin and were also dependent on Hap1 in all media, including the four genes (CFEM hemophores RBT5 and PGA7, and FRP1, FRP2) known to be essential for heme acquisition. Three genes that were induced by hemin but that were dependent on Hap1 only in the presence of hemin, include another CFEM hemophore, CSA1. A much larger group of genes (91, or 38.6%) was induced by hemin both in the wild-type and hap1 -/strains. GO analysis of this group of genes (Fig 6B; S4 Table, +He vs YPD, HAP1 \ hap1 -) revealed an enrichment in genes involved in the response to reactive oxygen species, and in carbohydrate transport, as seen also in the GO analysis described in panel A. In summary, hemin induces a response to reactive oxygen species and suppresses ribosome biosynthesis independently of HAP1, whereas HAP1 induces mainly genes of metal homeostasis, specifically of uptake and homeostasis of heme.
To obtain a more detailed view of genes dependent on Hap1, in Table 2 we have listed the genes most expressed (3x or more) in hemin-supplemented medium in the presence vs. the absence of HAP1. Strikingly, the top of the list is populated by the CFEM hemophores RBT5, PGA7 and CSA1, by FRP1 and FRP2, which are ferric reductase-like proteins required for heme uptake, and by the heme oxygenase HMX1. In Table 3, we have compared expression of these six genes under all four conditions (in HAP1 and hap1 -/-, with or without added hemin). What emerges from this comparison is that Hap1 is responsible not only for expression of these genes when hemin is added to the medium, but also for baseline expression in the absence of added hemin in the medium (compare HAP1 YPD to hap1 -/-YPD). Furthermore, in the absence of HAP1, exposure to hemin does not appear to induce these genes.  Table. https://doi.org/10.1371/journal.pgen.1010390.g006

PLOS GENETICS
We next asked whether genes induced or repressed in the HAP1 vs. the hap1 -/cells were enriched for the putative Hap1 consensus binding site CGGNNNTANCGG [37]. As mentioned above, in the haploid genome, this motif was found at 13 sites in intergenic regions, adjacent to 26 genes (0.4% of the total; S6 Table), including HAP1 itself. Out of 15 genes most induced by Hap1 (3x and more), the motif is found adjacent to 3 genes (20%) (p < 0.00003) ( Table 2). In contrast, the motif is not found adjacent to any of the 55 genes repressed 2x or more in HAP1 vs. hap1 -/cells. When a single mismatch is allowed, the motif is found adjacent to 1678 genes (28%) and in 11 of the 15 most-expressed genes (73%) (p < 0.0004) ( Table 2). We conclude that the binding motif of C. albicans Hap1 likely resembles that of S. cerevisiae Hap1, and that some of the most-highly induced genes in HAP1 vs. hap1 -/cells (Table 2) are likely to be direct targets of C. albicans Hap1.

Role of Hap1 in iron starvation-induced gene expression
The CFEM hemophore genes, as well as FRP1 and FRP2, were previously found to be among the most strongly induced genes under conditions of iron limitation [41,42]. Here however,

PLOS GENETICS
we have shown that their expression depends upon the heme-regulated Hap1 transcription factor. Thus, to clarify the role of Hap1 in expression of these genes under iron limitation, we measured expression of several of the Hap1 target genes, RBT5, PGA7, FRP1 and FRP2, as well as HAP1 itself, under conditions of iron limitation. To impose iron limitation, cells were transferred to medium containing the iron chelator ferrozine, which was previously shown to strongly induce RBT5 [16]. As shown in Fig 7A, all genes tested were strongly induced by the shift to ferrozine medium, except HAP1 itself, which showed only a limited induction at the earliest time-point. Kinetics of induction differed: FRP2, in particular, showed its greatest induction at early time points, compared to RBT5 and PGA7, which were gradually induced between the 2 h and the 6 h time point. Induction was also seen in the hap1 -/background, however in all genes tested, the extent of induction was reduced in the absence of Hap1, with the strongest effect detected with FRP2. In parallel we also followed the induction of the Rbt5 protein by Western blotting, and found that the amount of Rbt5 detectable upon ferrozine induction was reduced in the hap1 -/strain compared to the HAP1 strain, confirming a reduced expression of RBT5 in the hap1 -/strain at the protein level as well (Fig 7B). We conclude that while induction of the heme uptake genes by iron starvation does not depend on Hap1, the amplitude of induction was reduced, to different extents in different genes, in its absence.

Discussion
Organisms respond to their internal and external environment to maintain homeostasis and acquire nutrients. For microorganisms in the animal host, an important environmental factor is iron scarcity, which can be remediated by the acquisition of host heme. Heme, in addition to being itself an essential cellular factor, is also potentially toxic, and its cytoplasmic levels must therefore be closely controlled. We have shown here that C. albicans uses a single hemeresponsive transcription factor, Hap1, to both maintain cytoplasmic heme homeostasis and activate the utilization of external hemin as heme and iron sources.
S. cerevisiae Hap1 is bound and activated by heme [34]. However, rather than being involved in heme uptake like C. albicans Hap1, S. cerevisiae Hap1 plays a key role in the regulation of gene expression by oxygen [43,44]. Since heme is central to oxidative metabolism and since heme synthesis requires oxygen, the presence of heme is thought to serve as a barometer for the presence of oxygen in the cell [34,45]. Thus, C. albicans Hap1 has diverged in function from S. cerevisiae Hap1 by acquiring a distinct set of target genes. The two proteins do nonetheless appear to have maintained a similar consensus binding site, based on its occurrence in the promoters of the most Hap1-dependent genes ( Table 2), however direct binding of Hap1 to these sites remains to be shown experimentally. In addition, while it is not known whether, like S. cerevisiae Hap1, C. albicans Hap1 binds heme directly, two sequences having the consensus Hap1 heme regulatory motif K/R C P V/I [46] are found in the C. albicans Hap1 sequence at positions 288 and 888 (S1 and S3 Figs). Thus, one likely explanation for the stabilization of C. albicans Hap1 that we detected when hemin is added to the medium (Fig 5D) is that it is a consequence of direct binding of heme.
On the assumption that binding of heme stabilizes and activates C. albicans Hap1, and given that the HAP1 promoter contains a Hap1 consensus binding site, activation of Hap1 by heme would, in a positive feedback loop, activate its own expression as well as expression of its other target genes. Under normal conditions, when heme is endogenously synthesized, excess heme in the cytoplasm would cause Hap1 to activate the heme oxygenase HMX1, which would cause degradation of heme and restoration of homeostasis. Upon exposure of the cell to external heme, an initial influx of heme into the cytoplasm would activate the heme uptake cascade, leading to more heme influx. Our proposed model for the regulation of heme uptake and homeostasis by Hap1 thus consists of a series of interlocked feedback loops (Fig 8). We don't suggest that this model completely describes heme uptake and utilization; for example, unchecked activity of the heme uptake and degradation systems in the presence of heme might, once the cellular heme and iron needs are met, cause iron overload. Preventing this would require the activity of iron-responsive factors to shut off heme uptake genes under iron excess.
Global expression analysis gave insights into the effect of heme overload on the cells, as well as into the specific role of Hap1. GO analysis indicated an activation of genes for oxidant detoxification after exposure to a high concentration of heme, supporting the proposed toxic effect of heme as being due, at least in part, to the generation of reactive oxygen species. An additional general effect of heme exposure was suppression of ribosome biogenesis. This is consistent with previous observations in S. cerevisiae that metal stresses as well as other stresses cause suppression of ribosome synthesis genes [47,48]. The oxidant detoxification and suppression of ribosome biogenesis responses that we detected here were not dependent upon Hap1, but could be due to activation of general stress response pathways.
Comparison of gene expression with and without HAP1 identified a subset of genes involved in cellular iron homeostasis, specifically in heme uptake. Additionally and unexpectedly, even in the absence of added hemin to the medium, the hap1 -/strain showed reduced rRNA biogenesis, as well as an increase in genes for glycolysis, carbohydrate transport, and response to chemical stress /toxic substance. A possible explanation is that the absence of Hap1 leads to deregulation of heme homeostasis and to higher cellular heme concentrationsas was detected using the HS1 heme sensor (Fig 4A)-which in turn activates the cellular stress response. The increased carbohydrate transport could be related to the shift to glycolysis. It is not clear why the presence of heme should lead to a shift to glycolysis; one possibility is that by shifting away from cellular oxygen consumption and from generation of ROS by the respiratory chain, the cells could mitigate potential heme-catalyzed oxidative damage.
In contrast to the differences between HAP1 and hap1 -/cells in regular medium, much less enrichment in GO categories was detected in HAP1 vs hap1 -/cells exposed to 50 μM hemin. At such high hemin concentrations, both the wild-type and the mutant cells might experience similar heme overload-compare e.g. cytoplasmic heme in the two types of cells at 50 μM hemin in Fig 4A-, and therefore the general response to high hemin would be unaffected by the absence of HAP1.
The CFEM hemophore genes, as well as FRP1 and FRP2, shown here to be regulated by Hap1, had previously been shown to be among the most strongly induced genes under conditions of iron limitation [41,42]. This raised the question whether Hap1 is involved in iron limitation-dependent expression of these genes, in addition to mediating their hemin-induced gene expression. On the one hand, previous data had shown that the iron-responsive transcription factor CBF directly binds the FRP1 promoter and affects its expression [30]. On the other hand, the screen for GaPPIX-resistant mutants described above was performed under iron-limiting conditions in order to ensure expression of the heme uptake pathway, and nonetheless identified the HAP1 mutant, suggesting that Hap1 is required even under conditions of gene induction by iron limitation. Furthermore, absence of HAP1 led to reduced utilization of heme-iron (Figs 2 and S4), even though these experiments are also performed under iron starvation conditions. When we tested the induction of heme uptake genes in iron limited medium in the hap1 -/mutant, we found that absence of Hap1 somewhat reduced the

PLOS GENETICS
amplitude of induction of RBT5, PGA7 and FRP1, but strongly reduced the induction of FRP2. The effect on FRP2 explains why HAP1 was identified in a screen for resistance to GaPPIX, since FRP2 is essential for full GaPPIX toxicity [22]. However FRP2 plays a minor role at best in heme-iron utilization in standard medium [22]. Thus, either the partial reduction in expression of RBT5, PGA7 and FRP1 in the iron-limited hap1 -/strain is sufficient to explain the reduced heme-iron utilization, or other, unknown genes regulated by Hap1 participate in heme-iron utilization. Regarding the latter possibility, examination of the genes induced in the presence of Hap1 could reveal new factors involved in fungal heme-iron utilization.
The role of Hap1 identified here, to specifically promote uptake of external heme, underscores the importance of host heme as a micronutrient for animal parasites such as C. albicans. Furthermore, although external hemin was shown to be able to serve as source for cellular heme in an artificial heme auxotroph [14], the observation that heme uptake genes are activated via Hap1 even in the presence of sufficient iron in the medium, indicates that external hemin, when available, is actually a preferred heme source for this organism. Given the complexity of heme synthesis [49], this strategy makes metabolic sense, and may turn out to be more widespread than previously thought. External heme uptake must be complemented by a system for transporting the acquired heme from the endocytic system to the different cellular locations where it is required. Our knowledge of intracellular transport mechanisms of externally acquired heme-or, for that matter, of endogenously synthesized heme-is very limited [1,38,50]. Existence of a dedicated uptake system of external heme in C. albicans makes it a pertinent organism to probe these questions.

Transposon insertion identification
Single colonies were grown overnight in YPD medium, DNA was extracted as described [56], and subjected to the FPNI DNA amplification protocol [57] using Ds-specific primers (S7 Table, primers 1-6) as described [25]. PCR reactions were tested for the presence of bands on an agarose gel and the positive reactions (>90%) were cleaned using the GenElute kit (Sigma-Aldrich) and sequenced by Sanger sequencing. The sequence was used to determine the position and orientation of the Ds insertion in the C. albicans genome.

Growth assays
For hemoglobin utilization, overnight cultures grown in YPD were diluted in the morning into a series of two-fold dilutions of hemoglobin in YPD + ferrozine or BPS. Cells were inoculated in flat-bottomed 96-well plates at OD 600 = 0.00001, 150 μl per well. Plates were incubated at 30˚C on an orbital shaker at 60 rpm and growth was measured by optical density (OD 600 ) after 2 and 3 days with an ELISA reader. Cells were resuspended with a multi-pipettor before each reading. Each culture was done in triplicate. For MPP sensitivity, cells were inoculated in flat-bottomed 96-well plates in YPD + 1 mM ferrozine at OD 600 = 0.0001, 150 μl per well. Plates were incubated at 30˚C on an orbital shaker at 60 rpm and growth was measured by optical density (OD 600 ) after 2 days with an ELISA reader.

Heme sensor binding assays
The protocol was as described [14].

Glucose chase and Western blotting
For glucose chase of Hap1-6xMyc, cells were grown overnight in YEP + 2% raffinose at 30˚C, and diluted in the morning to OD 600 = 0.5, grown for 3 hours in YEP + 2% maltose for induction, with 20 μM hemin or without hemin, then 2% glucose was added and 1 ml aliquots were taken at 0', 30', 60' and 90'. Samples were then prepared as described [58]. Briefly, aliquots were treated with 0.25 N NaOH, 1% 2-mercaptoethanol, TCA-precipitated, washed in acetone and resuspended in protein loading buffer, then separated by SDS-PAGE and Western blotting. The blots were reacted with the 9E10 anti-Myc monoclonal antibody, and with monoclonal anti-tubulin (DM1A, Sigma) as internal control, and the bands were visualized by ECL and quantitated using an a ImageQuant LAS4010 apparatus. For detection of Rbt5, a previously described rabbit polyclonal antiserum [16] was used, and C4 anti-actin monoclonal antibody (MP Biomedicals) as internal control, and quantitated as above.

Microscopy
Cells were imaged with a Zeiss Axioskop Imager epifluorescence microscope equipped with DIC optics, using a 100X objective. For visualizing Hap1-GFP, the GFP filter set was used with an exposure time of 6 seconds. Nuclear stain: Hoechst 33342 was added to an exponentially growing culture to 10 μg/ml from a 1 mg/ml stock in water and the cells were incubated a further 30 min at 30˚C prior to microscopy visualization. For visualizing Hoechst 33342 by epifluorescence microscopy, the DAPI filter set was used with an exposure time of 1 second.

RNA extraction and analysis
For RT-PCR, for each sample 10 ml of culture was spun down and the pellet was frozen in liquid nitrogen and stored at -80˚C. The pellet was re-suspended in 400 μl lysis buffer (10 mM Tris-HCl pH 7.5, 10 mM EDTA, 0.5% SDS), and 400 μl acidic phenol, vortexed and incubated at 65˚C for 1 hour with occasional vortexing ("hot phenol" method [59]). The tubes were then placed on ice for 10 min and then centrifuged for 10 min at 4˚C, max speed. The upper layer was removed to a new tube, re-extracted with equal volume of phenol-chloroform 1:5 and vortexed. The tubes were centrifuged again for 10 min and the upper layer was removed into new tube, re-extracted with equal volume of chloroform and vortexed. The tubes were centrifuged again for 10 min and the upper layer was removed into new tube, for ethanol precipitation. Next, samples were treated with DNase to remove contaminating DNA from RNA using TURBO DNA-free Kit according to the manufacturer's instructions, followed by cDNA synthesis from 1000 ng of RNA, according to the Applied Biosystems High-Capacity RNA-to-cDNA Kit in total 20 μl reaction. RT-PCR (primers 15-28) was performed using QuantStudio 3 Real-Time PCR System equipment with Fast SYBR Green Master Mix in a final volume 10 μl. Expression in each sample was normalized to ACT1 expression.
For RNA-SEQ, mutant and reintegrant cells were grown overnight in YPD, diluted in the morning in YPD to OD 600 = 0.5. After 2.5 h at 30˚, the cultures were split in two and 50 μM hemin was added from a 2 mM stock, and samples were taken after 30 min. RNA was extracted using the "hot phenol" method [59]. Library construction and analysis was performed at the BCF Technion Genomics Center with NEBNext Ultra II Directional RNA Library Prep Kit for Illumina, cat no. E7760, using 800ng total RNA as a starting material. mRNAs pull-down was performed using the Magnetic Isolation Module (NEB, cat no. E7490). After construction, the concentration of each library was measured using Qubit (Invitrogen) and the size was determined using the TapeStation 4200 with the High Sensitivity D1000 kit (cat no. 5067-5584). All libraries were mixed into a single tube with equal molarity. The RNAseq data was generated on Illumina NextSeq2000, using P2 100 cycles (Read1-100; Index1-8; Index2-8) (Illumina, cat no. 20046811). Quality control was assessed using Fastqc (v0.11.8), reads were trimmed for adapters, low quality 3'and minimum length of 20 using CUTADAPT (v1.10). 100 bp single reads were aligned to a Candida albicans reference genome (GCF_000182965.3_ASM18296v3_genomic.fna from NCBI) using Tophat (v2.1.0, uses Bow-tie2 version 2.2.6). Between 26 x 10 6 and 31 x 10 6 uniquely mapped reads (~96% of total reads) were obtained per sample. The number of reads per gene was counted using HTSeq-count (v2.2.0). A statistical analysis was preformed using DESeq2 R package (v1.28.0) [60]. The number of reads per gene was extracted into merged_counts.csv and normalized_counts.csv files for raw counts and normalized counts, respectively. The similarity between samples was evaluated within DESeq2 package using correlation matrix, shown in two plots-heatmap and principal component analysis (PCA). Raw data are available at NCBI's GEO database under accession number GSE202577.
GO analysis on "Biological Process" was carried out using PANTHER [61] (http:// geneontology.org/), but using the list of 5921 protein-coding genes detected in our experiment as reference, rather than the total list of genes. Intersects between lists were obtained and Venn diagrams were drawn with Venny 2.1 (https://bioinfogp.cnb.csic.es/tools/venny/index.html).

Sequence comparison
The S. cerevisiae Hap1 and C. albicans Zcf20/Hap1 sequences were used to identify homologs in Saccharomycetales genomes using the Candida Genome Database (www.candidagenome. org) and Saccharomyces Genome Database (www.yeastgenome.org). 46 closest matches in 38 genomes (S1 Data) were aligned using the MAFFT G-INS-i algorithm [63], followed by tree building using the NJ method with 100 X resampling. Similar trees were obtained with other algorithms (UPGMA) and aligning methods (NCBI COBALT).