The ability to acquire iron from host tissues is a major virulence factor of pathogenic microorganisms. Candida albicans is an important fungal pathogen, responsible for an increasing proportion of systemic infections. C. albicans, like many pathogenic bacteria, is able to utilize haemin and haemoglobin as iron sources. However, the molecular basis of this pathway in pathogenic fungi is unknown. Here, we identify a conserved family of plasma membrane-anchored proteins as haem-binding proteins that are involved in haem-iron utilization. We isolated RBT51 as a gene that is sufficient by itself to confer to S. cerevisiae the ability to utilize haemoglobin iron. RBT51 is highly homologous to RBT5, which was previously identified as a gene negatively regulated by the transcriptional suppressor CaTup1. Rbt5 and Rbt51 are mannosylated proteins that carry the conserved CFEM domain. We find that RBT5 is strongly induced by starvation for iron, and that deletion of RBT5 is by itself sufficient to significantly reduce the ability of C. albicans to utilize haemin and haemoglobin as iron sources. Iron starvation-inducible, antigenically cross-reacting haem-binding proteins are also present in other Candida species that are able to utilize haem-iron, underscoring the conservation of this iron acquisition pathway among pathogenic fungi.
The dimorphic fungus Candida albicans is a commensal resident of the gastrointestinal tract that commonly causes mucosal, cutaneous or nail infections. Among immunocompromised or debilitated patients, C. albicans can cause life-threatening, disseminated infection (Richardson and Warnock, 1997). The ability to cause systemic infection depends on specific virulence factors that enable C. albicans to adhere to epithelial cells, penetrate tissues and cause damage (Hube et al., 1997; Lo et al., 1997; Sanglard et al., 1997; Gale et al., 1998). In addition, like all microbial pathogens, C. albicans needs to be able to extract nutrients from the host environment in order to proliferate. An important limiting nutrient in host tissues is iron: it has long been known that a number of host proteins play key defence roles against microorganisms by withholding free iron, a defence method that has been called ‘nutritional immunity’ (Kochan, 1973; Weinberg, 1975); conversely, successful pathogens have by necessity evolved mechanisms to acquire iron from the host tissues (Payne, 1993). Different mechanisms for extracting iron from host proteins have been characterized in bacterial pathogens, including utilization of haem and haemoglobin iron (Ratledge and Dover, 2000). The pathways of haemoglobin iron utilization were extensively characterized in Gram-negative bacteria (reviewed in Genco and Dixon, 2001), and a haem-iron acquisition apparatus was recently described in a Gram-positive pathogen (Dryla et al., 2003; Mazmanian et al., 2003).
Most of our information on fungal iron uptake mechanisms was derived from studied in the related, but non-pathogenic yeast Saccharomyces cerevisiae (recently reviewed by Van Ho et al., 2002; Kosman, 2003). S. cerevisiae low-affinity iron transport depends on the ferric reductases Fre1 and Fre2 (Georgatsou and Alexandraki, 1994) and the Fe2+ permease Fet4 (Dix et al., 1994; Dix et al., 1997). High-affinity iron import depends on the iron permease Ftr1 (Stearman et al., 1996) in conjunction with the cell-surface multicopper ferroxidase, Fet3 (Askwith et al., 1994; De Silva et al., 1995), which acts downstream of the reductases. Biogenesis of the Fet3 multicopper ferroxidase occurs in a trans-Golgi compartment and requires the activity of an intracellular copper transporter, Ccc2 (Yuan et al., 1995). Therefore, S. cerevisiae mutants lacking Ccc2 copper transport activity are defective in high-affinity iron import (Yuan et al., 1995). Aside from elemental iron, fungi are also able to utilize iron bound to siderophores. One mechanism of siderophore-iron utilization by S. cerevisiae consists in extracellular reduction of the siderophore-bound ferric iron to Fe2+ (Lesuisse and Labbe, 1989; Yun et al., 2001), which causes the ferrous ion to be released from the siderophore and taken up by the high-affinity iron uptake pathway. A second mechanism of siderophore–iron acquisition relies on specific transporters that import the whole siderophore–iron complex into the cell (Lesuisse et al., 1998; Heymann et al., 1999; Yun et al., 2000).
In C. albicans, components of the high-affinity iron import system have been identified based on homology with their S. cerevisiae counterparts, and/or by functional complementation of the corresponding S. cerevisiae mutants. These components include the iron permease CaFtr1 (Ramanan and Wang, 2000), the ferroxidase CaFet3 (Eck et al., 1999) and the copper transporter CaCcc2 (Weissman et al., 2002). A C. albicans siderophore transporter, CaArn1/CaSit1, was similarly identified by homology with its S. cerevisiae counterpart (Ardon et al., 2001; Heymann et al., 2002). Importantly, unlike S. cerevisiae, but like many microbial pathogens, C. albicans is also able to utilize haemoglobin and haemin as iron sources (Moors et al., 1992; Weissman et al., 2002; Santos et al., 2003). Haem-iron utilization by C. albicans does not depend on the high-affinity iron acquisition pathway (Weissman et al., 2002; Santos et al., 2003). In an attempt to identify components of the haem-iron utilization pathway in C. albicans, we performed a screen for genes able to confer the ability to utilize haemoglobin iron to S. cerevisiae. The genes we identified encode extracellular haem receptors that play a central role in the haem-iron utilization pathway.
RBT51 confers haemoglobin-iron utilization to S. cerevisiae
A S. cerevisiae ccc2Δ mutant, defective in high-affinity iron uptake (Yuan et al., 1995), and therefore unable to grow in the presence of the iron chelator ferrozine, was transformed with a C. albicans genomic library, and plated on YPD medium containing ferrozine and supplemented with haemoglobin. The genomic library plasmids that consistently enabled the S. cerevisiae cells to grow on this medium fell into two classes. One set of plasmids that enabled growth on ferrozine both in the presence and in the absence of haemoglobin carried the CaCCC2 gene, as expected (Weissman et al., 2002). A second set of plasmids that enabled weak but haemoglobin-dependent growth on ferrozine medium carried an open reading frame (orf 19.5674; http://www-sequence.stanford.edu/group/candida/) predicted to encode an extracellular glycosylphosphatidylinositol (GPI)-anchored protein (De Groot et al., 2003). The open reading frame displayed high homology to RBT5 (70% identity at the protein sequence level; Fig. 1A), a gene previously isolated as being repressed by the transcriptional repressor Tup1 (Braun et al., 2000); therefore, we called this gene RBT51. However, whereas the RBT51-containing plasmid was able to confer to S. cerevisiae the ability to utilize haemoglobin as an iron source, a plasmid containing the genomic region encompassing RBT5 was unable to do so (Fig. 1B). Rbt51 and Rbt5 belong to a family of at least five homologous C. albicans proteins, of which one, Wap1/Csa1, has been previously identified as an Rbt5 homologue (Braun et al., 2000) and a mycelial surface antigen (Lamarre et al., 2000). Figure 1C shows a homology tree of Rbt5, Rbt51, Wap1, and the two additional open reading frames with high homology to Rbt51. All five proteins contain the recently described CFEM domain (Kulkarni et al., 2003) (Wap1 contains four repeats of that domain), which has as main sequence feature a set of eight cysteines distributed with a characteristic spacing.
RBT5 is required for efficient haem-iron utilization in C. albicans
The ability to utilize haem-iron was monitored either in RPMI 1640 medium containing the human serum iron chelator apotransferrin or in YPD medium containing ferrozine, the latter medium requiring that the strain be defective in high-affinity iron acquisition (e.g. a Caccc2Δ mutant). We initially assessed the role of Candida albicans RBT51, RBT5 and WAP1 in YPD + ferrozine, the same medium that was used to isolate RBT51 in S. cerevisiae. For that purpose, deletions of these genes were introduced, singly and in combination, in the C. albicans ccc2Δ background. The starting Caccc2/Caccc2, rbt5/rbt5 and wap1/wap1 deletions have been described (Braun et al., 2000; Weissman et al., 2002); the rbt51/rbt51 deletion was ascertained by Southern blotting (Fig. 2A). The Caccc2Δ rbt5Δ strain displayed a significant decrease in the ability to utilize haemoglobin in YPD + ferrozine (Fig. 2B). However, in contrast with the results obtained in S. cerevisiae, where RBT51 but not RBT5 enabled growth on haemoglobin, the C. albicans strain deleted for RBT51 did not show any growth phenotype in the presence of haemoglobin as only iron source, nor did the C. albicans wap1 deletion (Fig. 2B). The rbt5Δ rbt51Δ double mutant, and rbt5Δ rbt51Δ wap1Δ triple mutant did not display significantly worse growth than the single rbt5Δ mutant (Fig. 2B). Because haemin utilization is inefficient in YPD, both haemin and haemoglobin utilization were tested in RPMI1640 + apotransferrin, and found to be defective in the rbt5Δ strain (Fig. 2C). This defect was largely, but not completely complemented by reintroducing the RBT5 gene in the mutant strain: the rbt5Δ/rbt5Δ <RBT5> strain grew slightly but consistently less well than the wild-type on either haemin or haemoglobin as sole iron source (Fig. 2C). However, a similar slight reduction in growth was also detected with the RBT5/rbt5Δ heterozygote, suggesting that this phenotype represents a haploinsufficiency of the strains carrying a single copy of RBT5. Taken together, these data indicate that at least under our experimental conditions, in C. albicans, Rbt5 plays the dominant role in haem-iron acquisition. Whether the two remaining Rbt5 homologues (Fig. 1C), or different proteins altogether, are responsible for the residual ability of the rbt5Δ mutant to utilize haem-iron remains to be investigated.
RBT5 is induced by iron starvation
An antiserum raised against recombinant Rbt51 was used to follow expression of the protein under different conditions by Western blotting analysis. As shown in Fig. 3A, in C. albicans wild-type cells, an immunoreactive polydispersed protein band visible at around 80 kDa was strongly induced by iron starvation. This band was equally prominent in iron-starved rbt51Δ cells, but was much weaker in the iron-starved rbt5Δ mutant, suggesting that in iron-starved cells, this induced signal represents Rbt5, rather than Rbt51. The cross-reactivity of Rbt5 with the anti-Rbt51 antibody is not surprising, in view of the 70% identity between the two proteins (Fig. 1A). The antibody is specific for Rbt5 and Rbt51, as evidenced by the observation that in the rbt5Δ rbt51Δ double mutant, either iron-starved or iron-sated, the 80 kDa polydispersed band is absent (note, however, the presence of an additional unrelated cross-reacting band migrating at around 50 kDa – indicated by a star throughout). Interestingly, under iron-replete conditions, the signal was weaker in the rbt51Δ mutant than in the rbt5Δ mutant, suggesting that in iron-sated cells, the level of Rbt51 protein is higher than that of Rbt5. Northern blot analysis of these strains with a probe hybridizing to the RBT5 and RBT51 sequences mirrored the protein analysis, indicating that induction of RBT5 by iron starvation occurs at the transcript level (Fig. 3B). In S. cerevisiae, the strain carrying a plasmid containing the RBT51 genomic region displayed a prominent immunoreactive band at around 80 kDa, whereas the strain carrying the RBT5 plasmid displayed a much weaker band (Fig. 3C). These observations might explain the difference between the activities of RBT5 and RBT51 in C. albicans versus S. cerevisiae as resulting from differential expression.
Rbt5 and Rbt51 are glycoproteins
The migration of Rbt5 and Rbt51 in SDS–polyacrylamide as polydispersed protein bands of apparent molecular weight of close to 80 kDa (Fig. 3) is highly aberrant relatively to the predicted molecular weight of these proteins, which is about 25 kDa. To test whether glycosylation could account for the slower migration of these proteins, extracts of C. albicans cell grown either in iron-replete medium or in ferrozine-supplemented medium were treated with NaOH, conditions that led to loss of sugar moieties via β-elimination. As shown in Fig. 4A, this treatment reduced the apparent molecular weight of anti-Rbt51-reactive protein (Rbt5 in the iron-starved cell extract, and possibly Rbt51 in the iron-replete cell extract) by 10–20 kDa. When the extracts were analysed on higher-density gels, no additional signal was detected below 50 kDa even after β-elimination (not shown). The 10–20 kDa decrease in apparent molecular weight most probably resulted from deglycosylation, although cleavage of the peptide chain at a single alkali-hypersensitive peptide bond could not initially be excluded as an alternative explanation.
Of the two common types of glycosylation, N-glycosylation usually occurs at the asparagine residue embedded in canonical N-X-S/T sequences (Marshall, 1972). These sequences are absent from either the Rbt5- or the Rbt51-predicted protein sequences. Furthermore, we failed to detect a change in apparent molecular weight of these proteins upon treatment with endoglycosidase H, which cleaves the core of N-linked oligosaccharides (data not shown). To test whether the other common type of glycosylation, O-mannosylation, could account for the aberrant migration of these proteins, Rbt51 was expressed in a set of S. cerevisiae mutants defective in this pathway. The PMT genes encode a family of partially redundant protein O-mannosyltransferases (reviewed in Strahl-Bolsinger et al., 1999), and GDA1 encodes a guanosine diphosphatase involved in the transport of GDP-mannose, a precursor of the biosynthesis of mannoproteins, into the Golgi lumen (Abeijon et al., 1993; Berninsone et al., 1994). As shown in Fig. 4B, two specific pmt mutants, pmt3 and pmt6, as well as gda1, displayed a significant acceleration in migration of Rbt51, suggesting that these proteins are involved in its post-translational modification. These results support deglycosylation as the reason for the decrease in apparent molecular weight of Rbt5 upon alkali treatment of C. albicans extracts (Fig. 4A), and furthermore, they point specifically to O-mannosylation as being responsible, at least in part, for the aberrant migration of Rbt5 and Rbt51.
Rbt5 is localized at the cell membrane
Rbt5 and Rbt51 carry a GPI consensus sequence at their C-termini, and a predicted signal sequence at their N-termini (De Groot et al., 2003). Therefore, these proteins were predicted to be located in the cell envelope, anchored either to the plasma membrane or to the cell wall. Consistent with this prediction, Rbt5 and Rbt51 are heavily glycosylated (Fig. 4), a modification typical of extracellular proteins. Extracellular localization was confirmed by fluorescently staining whole cells with the anti-Rbt51/Rbt5 antibody: an RBT5-dependent periplasmic signal was visible in unpermeabilized cells (Fig. 5A). We tested whether this signal represented plasma membrane or cell wall localization by using Western blotting and immunodetection to assay the distribution of Rbt5 in supernatant versus pellet of zymolyase-treated cells. As shown in Fig. 5B, Rbt5 was virtually exclusively found in the cell pellet, under digestion conditions that were sufficient to release significant amounts of a C. albicans cell wall protein, the 3H8-reactive protein (Marcilla et al., 1999), to the supernatant. Thus, Rbt5 is anchored to the cell membrane rather than to the cell wall.
Rbt5 and Rbt51 function as haem receptors
The localization of Rbt5 suggested it may function as a haem receptor. 55Fe-haemin bound to wild-type C. albicans cells in a saturatable fashion (data not shown; see also Santos et al., 2003), indicating the presence of a receptor. Furthermore, both cold haemin and haemoglobin could compete with binding of 55Fe-haemin, indicating that a single class of receptors bound both compounds (Fig. 6). In the presence of 1 µM 55Fe-haemin, the IC50 of haemoglobin was about 0.2 µM, indicating that the receptor has a higher affinity for haemoglobin. Total binding to C. albicans cells was induced some 20-fold after growth in iron-limited medium, from about 1.2 × 105 to about 3 × 106 molecules haemin/cell (Table 1). We next tested the binding of 55Fe-haemin to mutant versus wild-type cells after iron starvation. Whereas no significant reduction in 55Fe-haemin binding could be detected in the rbt51Δ and wap1Δ mutants compared to the wild-type, the rbt5Δ mutant showed threefold reduction of 55Fe-haemin binding, as did the rbt5Δ rbt51Δ double mutant. Deletion of WAP1 from the rbt5Δ rbt51Δ mutant led to a further significant reduction in 55Fe-haemin binding to about 15% of the wild-type cells. In addition, S. cerevisiae cells, which where normally unable to bind 55Fe-haemin, showed a significant amount of 55Fe-haemin binding upon expression of RBT51 (Table 1). These genetic results suggest that RBT5, RBT51 and WAP1 are all able to contribute to the haemin-binding capacity of the cell, the WAP1 contribution in C. albicans becoming significant only in the absence of RBT5. To demonstrate directly that Rbt51 and Rbt5 are haemin-binding proteins, both recombinant Rbt51 and an extract of iron-starved C. albicans cells were incubated with haemin-agarose. As shown in Fig. 7 (right), recombinant Rbt51 bound specifically to the haemin-agarose column. A protein from the yeast cell extract, migrating at ≈ 70 kDa, also bound specifically to the haemin-agarose column, and this binding could be out-competed by the addition of free haemin (Fig. 7, left). The ≈ 70 kDa band was absent from the rbt5Δ mutant cell extract, indicating that it represents Rbt5.
Table 1. Binding of 55Fe-haemin to the indicated C. albicans and S. cerevisiae strains.
Strain and growth medium
55Fe-haemin binding pmol per 107 cells ± SD
C. albicans wild-type, YPD
2.1 ± 0.3
C. albicans wild-type, YPD + ferrozine
48 ± 4
C. albicans wap1Δ, YPD + ferrozine
44 ± 2
C. albicans rbt51Δ, YPD + ferrozine
38 ± 3.5
C. albicans rbt5Δ, YPD + ferrozine
15 ± 1
C. albicans rbt5Δ rbt51Δ, YPD + ferrozine
15.9 ± 1
C. albicans rbt5Δ rbt51Δ wap1Δ, YPD + ferrozine
6.8 ± 1.6
S. cerevisiae <vector>, SC – URA
0.25 ± 0.1
S. cerevisiae <pADH1::RBT51>, SC – URA
2.35 ± 0.13
Conservation of the Rbt5 family of proteins
To test whether the Rbt5/Rbt51 haem-iron utilization pathway is conserved among other opportunistic pathogens of the Candida genus, we first tested the ability of different Candida species to utilize haemoglobin or haemin as iron sources. C. glabrata, C. parapsilosis, C. tropicalis and C. krusei are the non-albicans yeasts most frequently isolated in bloodstream infections (Pfaller et al., 2003). As shown in Fig. 8A, these four species varied in their ability to utilize haemoglobin and haemin as iron sources: C. tropicalis and C. parapsilosis were able to efficiently utilize either haemoglobin or haemin, whereas C. glabrata required markedly higher concentrations to grow, and C. krusei showed limited growth only at the highest haem-iron concentration. Western blotting analysis of a protein extract of these four species revealed, in C. tropicalis and C. parapsilosis, but not in the two latter, an anti-Rbt51 cross-reacting protein that was strongly induced by iron starvation (Fig. 8B). These Rbt51 cross-reacting proteins bound specifically to a haemin-agarose column (Fig. 8C). These data suggest that proteins homologous to Rbt51 may be involved in haem-iron utilization in other pathogenic yeasts as well. The disparity in apparent molecular weight between the C. tropicalis haemin-binding protein and Rbt5/51 could reflect differences in polypeptide chain length, but could also be explained by different levels of glycosylation – for example, Ccw14, an S. cerevisiae CFEM-domain mannoprotein with a predicted molecular weight of 24 kDa, similar to that of Rbt5 and Rbt51, migrates at a height of 160 kDa on SDS–PAGE (Moukadiri et al., 1997).
Many pathogenic microorganisms are able to utilize haemin or haemoglobin as iron sources. Best described to date are the mechanisms of haem-iron acquisition in Gram-negative pathogenic and commensal bacteria. These mechanisms include bacterial haemophores that take up haem from host haem proteins and deliver it to receptors on the bacterial outer membrane, or outer membrane receptors that bind directly to haemin or haem proteins, and then extract and transport the haem across the outer membrane into the periplasmic space, after which the haem is transported via permeases into the cytoplasm, to be finally degraded by haem oxygenases (reviewed in Wandersman and Stojiljkovic, 2000; Genco and Dixon, 2001). Recent studies that address Gram-positive haem-acquisition pathways have identified a family of haemin-, haemoglobin- and haptoglobin-haemoglobin-binding cell wall proteins, as well as a putative haem permease in Staphylococcus aureus (Dryla et al., 2003; Mazmanian et al., 2003).
In contrast to bacterial haem utilization pathways, the fungal pathways of haem-iron utilization have only recently started to be investigated. A C. albicans haem oxygenase, CaHmx1, that is necessary for the utilization of haemin as iron source, was isolated by homology with the S. cerevisiae protein (Santos et al., 2003), and was shown to metabolize haemin to α-biliverdin (Pendrak et al., 2004). In the present study, we identify C. albicans proteins, Rbt5 and Rbt51, involved in the haem-iron utilization and in the haemin-binding capacity of the cell, either when ectopically expressed in S. cerevisiae (Rbt51) or when deleted from C. albicans (Rbt5). Based on the high homology of these proteins, we assume that they fulfil a similar function, and that their differential effect in S. cerevisiae versus C. albicans mostly reflects their patttern of expression under our assay conditions. Sequence comparisons failed to reveal any homology between Rbt5 or Rbt51 and bacterial haem receptors; Rbt5 and Rbt51 do, however, contain the recently identified CFEM domain, a conserved sequence motif of unknown function, characterized mainly by a set of eight distinctly spaced cysteines (Kulkarni et al., 2003). CFEM proteins can be detected in the genomes of several pathogenic fungi, both in human pathogens such as C. immitis and H. capsulatum (a yeast which was independently shown to bind and utilize haemin; Foster, 2002) and in plant pathogens such as M. grisea and Fusarium sporotrichioides. Rbt5 and Rbt51 constitute the first CFEM proteins to be assigned a distinct function. However, we do not mean to imply that we expect all the CFEM proteins to function as haem-binding proteins. In fact, we believe this to be unlikely, based on the fact that S. cerevisiae, which in our assays is devoid of haem-binding capacity, contains a CFEM-domain mannoprotein, Ccw14 (Moukadiri et al., 1997). Future analysis of the deletion phenotype of CFEM domain proteins in other organisms, and the possible identification of the C. tropicalis and C. parapsilosis Rbt5 homologues once the genome sequence of these organisms becomes available, may enable us to identify a subset of CFEM domain sequences that are specific for haem-binding proteins.
We find that Rbt5 is an extracellular membrane protein. Based on its sequence, Rbt5 – like Rbt51 – is predicted to be attached to the membrane via a GPI anchor (De Groot et al., 2003). An Rbt5–green fluorescent protein (GFP) fusion was indeed found to be released from a membrane fraction with a phosphatidylinositol-specific phospholipase C, as would be expected from GPI-anchored protein (Mao et al., 2003); however, that study also found that a significant fraction of the Rbt5-GFP fusion protein was cross-linked to the cell wall glucans. In contrast, we found that native Rbt5 was exclusively located in the cell membrane. This discrepancy could result from the different growth conditions used in the two sets of experiments, or alternatively, to the fact that Mao et al.'s Rbt5–GFP fusion only contains the N- and C-terminal sequences of Rbt5, residues 43–199 of Rbt5 having been substituted with the GFP sequence, which may have caused in some way a partial mislocalization of the protein.
Our observations along with those of Santos et al. (2003) allow us to sketch the outlines of the haem-iron utilization pathway in Candida spp. The requirement in C. albicans of an intracellular haem oxygenase for haem-iron utilization (Santos et al., 2003) implies that the iron atom is extracted from the haem after import into the cell. In accordance with this notion, the high-affinity iron import pathway does not appear to be required for haem-iron acquisition, because both Caccc2 mutants (Fig. 1; Weissman et al., 2002) and Caftr1 mutants (Santos et al., 2003 and our unpubl. results) can efficiently utilize haemin and haemoglobin as iron sources. Thus, we suggest that the role of the Rbt5/Rbt51 family of proteins is to facilitate internalization of the haem. One possible mechanism would be for Rbt5/Rbt51 to tether the haem at the cell surface, either to facilitate its diffusion across the plasma membrane (Genco and Dixon, 2001) or to allow uptake by a yet-to-be-discovered haem transporter. Alternatively, the haem–Rbt5/51 complex could be internalized, followed by release of the haem to the cytoplasm.
Haemoglobin-iron utilization requires an initial step of extraction of the haem from the globin. The first mechanism for haem internalization suggested above would require that this extraction be carried out extracellularly, whereas the second mechanism would be compatible with an internalization of the haemoglobin with the receptor, followed by extraction of the haem from the globin within the cell (presumably, concomitant with degradation of the globin polypeptide chains). Haemoglobin has been previously shown to bind to a saturatable receptor on the C. albicans surface, and to induce the appearance of a fibronectin receptor (Pendrak et al., 2000). Whether or not the Rbt5/Rbt51 class of proteins corresponds to the haemoglobin receptor involved in the induction of the fibronectin receptor remains to be investigated. However, Pendrak et al. (2000) also made the interesting observation, which we could confirm (data not shown), that in the presence of metabolically active C. albicans cells, haemoglobin rapidly precipitated to form aggregates, and suggested that this aggregate formation could have resulted from a haem extraction process. This possibility implies that the haem moiety is in fact extracted extracellularly from the haemoglobin, followed by uptake of the haem, but leaves unresolved whether subsequently the haem is internalized together with the Rbt5/51 proteins or is taken up via a different mechanism.
The RBT5 gene, under our in vitro assay conditions, plays the dominant role in haem-iron utilization in Candida, and is strongly induced by iron limitation. RBT5 was previously identified as a gene repressed by the transcriptional repressor Tup1 (Braun et al., 2000). Interestingly, reductive iron uptake genes (Knight et al., 2002), as well as the siderophore transporter gene CaSIT1 (Lesuisse et al., 2002), were found to be derepressed in the Catup1Δ background. Thus, it would appear that every single iron acquisition pathway identified thus far is deregulated in the Catup1Δ mutant. Although the Catup1Δ phenotypes are too pleiotropic, and its known targets too varied, to be explained by a deregulation of iron uptake alone (Braun and Johnson, 1997; Braun et al., 2000), a likely explanation is that an iron satiety-specific transcriptional regulator uses Tup1 as co-repressor.
By analogy with bacterial pathogens (Stojiljkovic et al., 1995; Torres et al., 2001), it is likely that the ability of C. albicans to utilize haem-iron enhances its virulence as a systemic pathogen, by enabling it to acquire iron from serum haemoglobin. Although there is very little free haemoglobin in serum, the haemolytic function of C. albicans (Manns et al., 1994) would increase its availability in the vicinity of the cell. However, the rbt5Δ mutant was previously reported to be as virulent as the wild-type strain in a mouse model of systemic infection (Braun et al., 2000). This result could still be reconciled with a role for the Rbt5/Rbt51 pathway in this specific experimental model of pathogenicity by several considerations. First, although we saw that RBT5 plays the dominant role in haem-iron utilization in vitro, the rbt5Δ mutant still retained a significant ability to utilize haemoglobin as an iron source (Fig. 2). This residual haem-iron utilization capacity, possibly because of other members of the Rbt5/Rbt51 family, may be sufficient to support growth in the animal. Second, it is possible that under the different conditions encountered by the invading microorganism during the course of systemic infection, RBT51 or other members of this gene family assume the dominant role in haem-iron utilization. Systematic deletions of all the members of this gene family, alone or in combination, will be required to address this question. The question whether haem-iron utilization is in fact required for systemic disease, at least in the mouse system, would best be addressed using a mutant completely devoid of haem-iron utilization capacity, such as the Cahmx1Δ mutant (Santos et al., 2003). On balance, we believe that the requirement for an iron acquisition capacity in the course of microbial infection, together with the conservation of the Rbt5/51 haem-binding proteins among several pathogenic yeasts, do suggest that this family of proteins contributes to virulence. Thus, the conserved Rbt5/Rbt51 family of haem-binding proteins as a whole, by virtue of their extracellular location, apparent conservation among several opportunistic pathogens of the Candida genus, and lack of homology to known human proteins, may constitute an attractive target for the development of specific inhibitors that would contribute to our limited armamentarium against fungal infections.
Strains and plasmids
Candida albicans strains are listed in Table 2. Gene deletions were performed sequentially using Fonzi and Irwin's Candida URA3 gene blaster (Fonzi and Irwin, 1993), checked by polymerase chain reaction (PCR) and confirmed by Southern blotting. The S. cerevisiae ccc2Δ strain and the Candida albicans genomic library have been described before (Liu et al., 1994; Gaxiola et al., 1998). The S. cerevisiae pmt1-3, 6, and gda1 deletion strains were generated by the S. cerevisiae gene deletion consortium and obtained via EUROSCARF (Frankfurt). The C. tropicalis and C. krusei strains used are type cultures from the Centraal Bureau voor Schimmelcultuur (Utrecht). C. parapsilosis and C. glabrata were obtained from Israela Berdicevski (Technion Faculty of Medicine). All four strains were reconfirmed using an API32 sugar assimilation diagnostic kit (Bio-Merieux).
RBT51 was first isolated as a clone from a C. albicans genomic library that extended between two Sau3A sites, from position −3366 to position +2771 relative to the RBT51 start codon. This fragment contains the complete RBT51 open reading frame, and two additional partial open reading frames. Plasmid pRBT51 (KB1189) was generated by introducing a ClaI-BamHI fragment extending from positions −790 to +1466 relative to the RBT51 start codon, into a 2 µURA3 vector, B2205. This plasmid contains only the RBT51 open reading frame. Plasmid pRBT5 (KB1302), containing the genomic region of RBT5 from Candida, was constructed by introducing a HindIII-KpnI PCR fragment, extending from position −550 to +1050 relative to the RBT5 start codon, into a 2 µURA3 vector, B2205. Plasmid KB1287 contains the same fragment cloned in the C. albicans integrative vector BES116 (Feng et al., 1999). The pADH1::RBT51 plasmid (KB1266) was generated by introducing the RBT51 coding sequence (positions −20 to +778) as an EcoRI-XhoI PCR fragment under the ADH1 promoter of plasmid KB1109; KB1109 contains the ADH1 promoter on a 1.5 kb EcoRI-BamHI fragment cloned in p416-GAL1 (Mumberg et al., 1994).
Yeast growth media (YPD: yeast extract peptone dextrose; SC: synthetic complete) were as described (Sherman et al., 1986). Growth curves versus concentrations of haemin and haemoglobin were obtained by diluting overnight cultures to an OD600 of 0.002 into a series of twofold dilutions of haemin or haemoglobin. The cultures were grown at 30°C in flat-bottomed 96-well plates and the optical density was read after 2 or 3 days using an ELISA reader.
The coding sequence of Rbt51 extending from codon 23 (at the predicted signal sequence cleavage site) to codon 235 (before the predicted GPI-anchoring site) was cloned as an NcoI-XhoI PCR fragment under the bacterial signal sequence and polyhistidine tag of pET25b (Novagen). Expression of the recombinant protein from the T7 promoter of pET25b was induced by the addition of 1 mM IPTG to a 500 ml bacterial culture, OD600 = 0.7; after 90 min, the cells were centrifuged and the pellet was resuspended in 80 ml of ice-cold 30 mM Tris pH 7.5, 1 mM EDTA, 20% sucrose. After 10 min on ice, the cells were centrifuged, and the cell pellet was resuspended in an ice-cold solution of 5 mM MgSO4. Periplasmic proteins released by this osmotic shock were then separated from the cells by an additional centrifugation. His6-rRbt51 was purified from the supernatant by binding to a nickel-agarose column in 5 mM MgSO4, 10 mM Tris pH 7.5, 10 mM imidazole, and eluted in 100 mM imidazole. This purified protein was used to raise a rabbit polyclonal antiserum. To show binding of rRbt51 to haemin-agarose, crude bacterial supernatant was directly loaded onto the haemin-agarose column.
Yeast protein extracts
To prepare total protein extracts, the yeast cells were incubated on ice for 10 min in 250 mM NaOH, 1% 2-mercaptoethanol, then trichloroacetic acid was added to 5%. The extract was centrifuged at 31 000 g for 10 min, the pellet was washed in ice-cold acetone, dried in a speed-vac centrifuge, and the cell pellet was resuspended and heated at 95°C in SDS–PAGE loading buffer containing 2% 2-mercaptoethanol. For β-elimination, incubation in 250 mM NaOH was carried out for 1 h at room temperature (longer incubation times caused loss of the protein signal on Western analysis, possibly because of degradation of the peptide backbone). Native protein extracts for the haemin column binding experiments were obtained by mechanical breakage of the cells with glass beads in binding buffer: 1% Triton X-100, 50 mM Tris pH 7.5, 1 M NaCl, 5 mM EDTA, and antiprotease cocktail (Sigma). Glass beads and insoluble material were removed by a low-speed centrifugation (1000 g, 10 min) followed by a high-speed centrifugation (31 000 g, 90 min). For binding to the haemin-agarose column or to the control glutathione-agarose column (both from Sigma), the cell extracts or recombinant Rbt51 were incubated with 20 µl of the appropriate column, first 15 min at 37°C, then overnight at 4°C in binding buffer, then the columns were washed three times with 1 ml of the binding buffer, and finally the bound proteins were eluted by heating in SDS–PAGE loading buffer. For the competition with free haemin, 2 mM haemin was added to the binding buffer.
55Fe-haemin synthesis and binding assays
55Fe (as 55FeCl3; NEN) was incorporated into protoporphyrin IX (Sigma) according to the method of Galbraith (Galbraith et al., 1985). The concentration of 55Fe-haemin was calculated from absorbance at 410 nm, and confirmed by competition with cold haemin (see Fig. 6). For measuring binding of 55Fe-haemin to yeast cells, the cells were incubated with 55Fe-haemin for 5 min in PBS + 0.05% Tween 20 (Sigma), centrifuged for 20 s in a microcentrifuge, washed once with the binding buffer, and finally the radioactivity remaining in the cell pellet was measured by scintillation counting. All binding assays were performed in triplicate. Non-specific background radioactivity values were obtained by incubating the cells in the presence of an additional 0.1 mM (100-fold excess) cold haemin, and substracted from the total bound radioactivity values. In preliminary experiments, no significant difference was found between binding at 4°C versus room temperature, indicating that under our assay conditions, binding rather than metabolism of 55Fe-haemin is monitored. We therefore did all our experiments at room temperature. Preliminary experiments also established that binding of 55Fe-haemin to the cell reached saturation at 0.5 µM. We therefore used a 55Fe-haemin concentration of 1 µM throughout our experiments, and up to 3 × 107 cells per assay. Under these conditions, up to 15% of total 55Fe-haemin was bound to the cells.
We thank Burk Braun, Israela Berdicevski and Yue Wang for yeast strains, Rafael Sentandreu for the 3H8 monoclonal antibody and Joseph Frey for advice with the 55Fe-haemin synthesis. Sequencing of Candida albicans at the Stanford Genome Technology Center was accomplished with the support of the NIDR and the Burroughs Wellcome Fund. Our work is funded by the Chief Scientist's Office, Israeli Ministry of Health and by the Wolfson Charitable Fund Center of Excellence for Studying Turnover of Proteins.