The pathogenic fungus, Candida albicans contains homologues of the transcriptional repressors ScTup1, ScMig1 and ScNrg1 found in budding yeast. In Saccharomyces cerevisiae, ScMig1 targets the ScTup1/ScSsn6 complex to the promoters of glucose repressed genes to repress their transcription. ScNrg1 is thought to act in a similar manner at other promoters. We have examined the roles of their homologues in C. albicans by transcript profiling with an array containing 2002 genes, representing about one quarter of the predicted number of open reading frames (ORFs) in C. albicans. The data revealed that CaNrg1 and CaTup1 regulate a different set of C. albicans genes from CaMig1 and CaTup1. This is consistent with the idea that CaMig1 and CaNrg1 target the CaTup1 repressor to specific subsets of C. albicans genes. However, CaMig1 and CaNrg1 repress other C. albicans genes in a CaTup1-independent fashion. The targets of CaMig1 and CaNrg1 repression, and phenotypic analyses of nrg1/nrg1 and mig1/mig1 mutants, indicate that these factors play differential roles in the regulation of metabolism, cellular morphogenesis and stress responses. Hence, the data provide important information both about the modes of action of these transcriptional regulators and their cellular roles. The transcript profiling data are available at http://www.pasteur.fr/recherche/unites/RIF/transcriptdata/.
Candida albicans is the major fungal pathogen in humans (Odds, 1988). It is carried as a relatively harmless commensal by about half of the population. However, when the balance between host and pathogen is disturbed, this fungus causes frequent and recurrent infections of the oral, gastrointestinal and urogenital tracts. In severely immunocompromised patients, C. albicans can infect the bloodstream and invade internal organs, often leading to mortality.
The mode of action of CaTup1 in C. albicans might be similar to that of ScTup1 in S. cerevisiae, suggesting that CaNrg1 and CaMig1 might target a CaTup1 repressor complex to distinct sets of C. albicans genes to repress their expression. The observations presented in this paper strongly support and extend this model. Our transcript profiling data reveal subsets of C. albicans genes that are regulated by CaMig1, CaNrg1 and CaTup1. The data also provide new insights into the regulatory functions of CaTup1, CaNrg1 and CaMig1, as well as the differential regulation secreted aspartyl proteinase (SAP) and agglutinin-like sequence (ALS) gene family members.
Generation of C. albicans arrays
Candida albicans genome sequence data from GenBank and from the Stanford's DNA Sequencing and Technology Center (Assembly 3; http://www-sequence.stanford.edu/group/candida/Tzung et al., 2001) were used to make gene arrays. Using these data, 3313 putative ORFs were identified, from which 2016 ORFs were selected for array construction. They comprised 364 GenBank entries, 1020 S cerevisiae homologues and 632 hypothetical ORFs. A 3′-fragment of 300–400 bp from each ORF was polymerase chain reaction (PCR)-amplified, yielding 2002 products that were arrayed on nylon membranes (Fig. 1). The array covers about one quarter of the expected ≈9000 protein coding genes in C. albicans. Further information on these ORFs is available at http://www.pasteur.fr/recherche/unites/RIF/.
A random set of 98 C. albicans ATCC66369 cDNA clones were partially sequenced to test the validity of our gene annotation (V. Marchais and J. Cottin, unpublished). To examine whether the 3′-ends of ORFs had been annotated correctly, we needed to match cDNA sequences to annotated ORFs for which downstream genome sequence information was available. However, eight out of the 98 cDNAs had no match within Assembly 3. This was consistent with the estimated genome coverage of the Assembly 3 sequence contigs (approximately 80%). Also, 34 out of the 98 clones matched Assembly 3 contigs that were not included in our annotation, suggesting that we had annotated about two-thirds of expressed C. albicans ORFs (see Experimental prodecures). Of the remaining 56 cDNAs, 14 originated from regions that were not annotated because no ORF was predicted. Presumably these reflect errors in Assembly 3 sequences, which were based on 1.7× coverage. The last 42 cDNAs matched predicted ORFs that were annotated. However, for six out of these 42 cDNAs, the region downstream of the corresponding gene was missing from the relevant Assembly 3 contig. Hence, 36 cDNAs corresponded to ORFs that we had annotated and for which downstream genome sequence was available. These 36 cDNAs could be used to validate our annotation. For 31 out of these 36 cDNAs (about 85%), the poly(A) tail started downstream of the predicted stop codon, as expected. For the remaining five cDNAs, the poly(A) tail started upstream from the predicted stop codon, suggesting further undetected frameshifts. Therefore, we estimate that our annotation was acceptable for about 85% of genes analysed.
The quality of the C. albicans arrays was examined by performing replicate hybridizations with the wild-type control strain, CAF2–1. A linear response was observed for dilutions of C. albicans genomic DNA spotted onto the membranes (Fig. 2A), confirming that the hybridizations were quantitative. Duplicate spots on a single membrane gave similar signals (Fig. 1), and replicate hybridizations showed correlation coefficients of > 0.93 (see Experimental procedures;Fig. 2B), confirming that the data were reproducible. Most blank spots and heterologous S. cerevisiae and Escherichia coli genes gave insignificant signals (Fig. 2B). Hence, the membranes were of high quality and gave reproducible signals.
Global expression levels in C. albicans
Expression was detected for 1851 (93%) of the 2002 C. albicans genes analysed. A similar proportion of S. cerevisiae genes were expressed (90%: Wodicka et al., 1997). Expressed genes were classified according to their expression level in the wild-type C. albicans strain (CAF2–1) relative to the actin gene. The ACT1 mRNA is often used as a standard against which the expression of other genes is measured (e.g. Planta et al., 1999; Brown et al., 2001). A small proportion of expressed genes (9%) were expressed at high levels (MX > 0.5 × MACT1: where MX = mean signal for gene X). This subset included housekeeping functions such as ribosomal proteins, members of the Hsp70 family of chaperones and factors involved in energy generation. About 30% of the genes were expressed at moderate levels (0.1 × MACT1 < MX < 0.5 × MACT1), and 61% were expressed at low levels (MX < 0.1 × MACT1). Therefore, a high proportion of C. albicans genes are expressed at low levels under the conditions analysed, and this is also the case in S. cerevisiae (Wodicka et al., 1997; Planta et al., 1999).
CaMig1, CaNrg1 and CaTup1 act mainly as repressors in C. albicans
We predicted that CaMig1, CaNrg1 and CaTup1 might operate in C. albicans in a similar fashion to their counterparts in S. cerevisiae. It follows from this model that CaMig1, CaNrg1 and CaTup1 all act as repressors in C. albicans. To test this, transcript profiling was performed on C. albicans tup1/tup1, mig1/mig1 and nrg1/nrg1 cells cultivated under conditions that normally promote growth of C. albicans in the yeast form, but result in filamentous growth for the tup1/tup1 and nrg1/nrg1 mutants. Inactivation of CaTup1, CaMig1 or CaNrg1 caused 9–13% of the C. albicans genes to be derepressed more than threefold, whereas the expression of only a small proportion of the genes was reduced more than threefold (Fig. 3). Therefore, as predicted, CaTup1, CaMig1 and CaNrg1 act primarily as negative regulators in C. albicans.
Some C. albicans genes (2.0%) were repressed more than threefold by the tup1/tup1 mutation, and significantly fewer genes were repressed in the mig1/mig1 and nrg1/nrg1 mutants (0.8% and 0.6% respectively). Therefore, CaTup1 also appears to activate the expression of some C. albicans genes.
Subsets of C. albicans genes are regulated by CaMig1-CaTup1, or CaNrg1-CaTup1
A second prediction of the model is that some C. albicans genes should be regulated by CaMig1 and CaTup1, whereas others should be regulated by CaNrg1 and CaTup1. As expected, some genes were derepressed in the mig1/mig1 and tup1/tup1 mutants, but were not derepressed in the nrg1/nrg1 strain (Fig. 4). Other C. albicans genes were derepressed in nrg1/nrg1 and tup1/tup1 cells, but not in mig1/mig1 cells. These data are consistent with the idea that CaNrg1 and CaMig1 target CaTup1 to distinct sets of genes.
Relationship between CaMig1-, CaNrg1- and CaTup1-regulated genes in C. albicans.
According to our initial model, CaMig1 targets CaTup1 to one set of C. albicans genes, and CaNrg1 targets CaTup1 to a different set of genes. This implied that there is no overlap between the subsets of CaMig1- and CaNrg1-regulated genes in C. albicans, but the situation is more complex. Some genes were derepressed in all three mutants (Fig. 5), suggesting a degree of overlap between the targets of CaMig1-CaTup1 and CaNrg1-CaTup1 regulation.
The model also implied that CaTup1 regulates all C. albicans genes that are regulated by CaMig1 and CaNrg1. However, a number of CaNrg1- and CaMig1-regulated genes were not derepressed in the tup1/tup1 mutant (Fig. 5). Hence, CaMig1 and CaNrg1 can act independently of CaTup1 to repress some C. albicans promoters. Therefore, CaMig1, CaNrg1 and CaTup1 make different contributions to several distinct regulatory circuits in C. albicans.
Cognate DNA binding sites in the promoters of CaMig1- and CaNrg1-regulated genes
ScMig1 interacts specifically with the DNA sequence 5′-(G/C)(C/T)GG(G/A)G (Nehlin and Ronne, 1990). CaMig1 is a functional homologue of ScMig1 (Zaragoza et al., 2000). Also, CaMig1 has been shown to be a DNA binding protein that interacts with the same region of the ScFBP1 promoter as ScMig1 (Zaragoza et al., 2000). Therefore, CaMig1 and ScMig1 appear to bind related DNA sequences. Hence, we scanned for the sequence 5′-(G/C)(C/T)GG(G/A)G in the promoters of the 25 C. albicans genes that are most strongly regulated by CaMig1 and CaTup1 (Table 1). Nineteen out of the 24 genes (79%), for which promoter sequence was available, carried this putative CaMig1 binding site, and many of these promoters carried multiple copies of this element. Also CaPCK1, which was not included on the array but which is repressed by CaMig1 (Murad et al., 2001), contains this element in its promoter region. However, some CaMig1 regulated promoters did not have this element, suggesting that they may be regulated indirectly by this factor. Also, some genes contained this element but were not derepressed in the mig1/mig1 strain, possibly because they are regulated by additional factors.
Table 1. Top 25 Mig1-Tup1-regulated C. albicans genes revealed by transcript profiling.
Fold regulation for a specific gene = (normalized mean signal in the mutant)/(normalized mean signal in the wild type). Nrg1-Tup1-regulated genes were defined as genes that displayed > threefold derepression in the nrg1/nrg1 and tup1/tup1 mutants, but < threefold derepression in the mig1/mig1 mutant.
A similar observation was made for CaNrg1. ScNrg1 binds the DNA sequence 5′-C4T or 5′-C3TC (Park et al., 1999), whereas CaNrg1 represses transcription via a related sequence element (A/C)(A/C/G)C3T [the Nrg1-responsive element (NRE): Murad et al., 2001]. Therefore, we searched for this element in the promoters of the 25 genes that are most strongly regulated by CaNrg1 and CaTup1. Nineteen (76%) out of these genes carried a NRE (Murad et al., 2001). Again, some CaNrg1-regulated promoters did not carry an NRE, suggesting that CaNrg1 might act indirectly upon these genes. Also, some NRE-containing genes were not derepressed in nrg1/nrg1 cells, possibly because they are repressed by other factors. However, site-directed mutagenesis of the two NREs in the ALS8 promoter released it from CaNrg1-mediated repression (Murad et al., 2001), confirming the central role of this element in CaNrg1 function. These data indicate that both CaMig1 and CaNrg1 act directly upon the promoters of most genes they regulate.
Cellular functions of CaMig1-, CaNrg1- and CaTup1-regulated genes
Our next objective was to examine the cellular roles of these regulatory circuits in C. albicans. Although the functions of only a small proportion of C. albicans genes have been examined experimentally, the functions of others (> 32%) may be inferred on the basis of their sequence similarity to S. cerevisiae genes.
Many of the C. albicans genes that were most strongly co-repressed by CaMig1 and CaTup1 are thought to execute metabolic functions in carbon source uptake and catabolism (Table 1; Fig. 5). They included a high-affinity glucose transporter (HGT1), a glucoamylase (GCA1), a galactokinase (GAL1) and a glycogen phosphorylase (GPH1), as well as alcohol, succinate semialdehyde and malate dehydrogenases (ADH1, UGA2, MDH1). In contrast, hypha-specific functions were concentrated in the subset of genes that were most strongly regulated by CaNrg1 and CaTup1 (Fig. 5). This list included HWP1, ALS3, ALS8 and ECE1, all of which are hypha-specific (Birse et al., 1993; Staab et al., 1996; Hoyer et al., 1998a; Brown, 2001).
Functional categories were assigned to those C. albicans genes that have a known function or a S. cerevisiae homologue of known function. For the purposes of this analysis, we assumed that a C. albicans ORF belongs to the same functional category as its S. cerevisiae counterpart (http://www.mips.biochem.mpg.de/proj/yeast). Interestingly, genes involved in metabolism and transport facilitation were significantly over-represented in the CaMig1-, CaNrg1- and CaTup1-regulated gene sets, compared with the regulated gene set as a whole (Fig. 6). In contrast, genes involved in protein destination, cell growth, cell division and DNA synthesis were relatively under-represented in these gene sets. Some degree of specialization was observed for CaMig1- and CaNrg1-regulated genes. Genes involved in cell rescue, defence, cell death and ageing were enriched in the CaNrg1- and CaTup1-regulated gene sets, whereas genes involved in energy generation were enriched in the CaMig1 and CaTup1 gene sets. Hence, CaMig1, CaNrg1 and CaTup1 appear to contribute differentially to the regulation of specific cellular processes.
Phenotypes of mig1/mig1, nrg1/nrg1 and tup1/tup1 mutants
One might expect mig1/mig1, nrg1/nrg1 and tup1/tup1 mutants to display phenotypes that reflect the regulatory targets of CaMig1, CaNrg1 and CaTup1 identified by transcript profiling. The targets of all three regulators were enriched in metabolic functions (Fig. 6). Consistent with this, mig1/mig1, nrg1/nrg1 and tup1/tup1 cells grew more slowly than wild-type cells (Table 2). The transcript profiling data suggested that CaMig1 might regulate genes involved in the utilization of non-preferred carbon sources (Fig. 5), as its counterpart in S. cerevisiae (Gancedo, 1998). Saccharomyces cerevisiae mig1 mutants are resistant to catabolite repression by 2-deoxyglucose, but it was not possible to test this phenotype in C. albicans because wild-type CAF-2 cells were resistant to this glucose analogue (not shown). However, the gluconeogenic CaPCK1 mRNA was derepressed in mig1/mig1 cells, but not in nrg1/nrg1 cells (Murad et al., 2001), confirming that CaMig1 plays a role in the regulation of carbon metabolism, as suggested by the transcript profiling data. Also, mig1/mig1 cells utilized oxygen at a significantly reduced rate (66%; 118 ± 13 nmol O2 min−1 OD−1) compared with wild-type and nrg1/nrg1 cells (178 ± 2.9 and 179 ± 2.9 nmol O2 min−1 OD−1 respectively). This is consistent with a role for CaMig1 in the regulation of energy-related functions (Fig. 6).
Table 2. Effects of mig1/mig1, nrg1/nrg1 and tup1/tup1 mutations upon the sensitivity of C. albicans to certain stresses.
The doubling time (min) in YPD for each strain is shown in parentheses. The effect of each stress upon the doubling time of a particular strain is expressed as a percentage of its doubling time in YPD.
YPD + 10 mM 3-AT
YPD + 0.7 M NaCl
YPD + 4.4 mM H2O2
YPD + 5% ethanol
Most known hypha-specific genes were present in the set of CaNrg1 and CaTup1 targets (Fig. 5), and the phenotypes of nrg1/nrg1 and tup1/tup1 mutants were consistent with this finding. Nrg1/nrg1 and tup1/tup1 cells, but not mig1/mig1 cells, displayed constitutive filamentous and invasive growth (Murad et al., 2001). CaNrg1 and CaTup1 targets were also enriched in cell rescue functions (Fig. 6), and nrg1/nrg1 and tup1/tup1 mutants displayed defects in some specific stress responses. These cells were more sensitive to H2O2 and ethanol than wild-type cells, but were equally resistant to high salt or the histidine analogue, 3-aminotriazole (Table 2). Therefore, there is a good correlation between the transcript profiling data and the phenotypes of mig1/mig1, nrg1/nrg1 and tup1/tup1 mutants. Clearly, CaMig1, CaNrg1 and CaTup1 contribute differentially to the regulation of specific cellular processes.
Regulation of SAP and ALS gene families
We analysed the expression of members of the ALS and SAP gene families in mig1/mig1, nrg1/nrg1 and tup1/tup1 cells, because of their important roles in C. albicans(Table 3). Most SAP genes were expressed at relatively low levels under the conditions analysed [growth in yeast extract peptone dextrose (YPD)]. This is consistent with their repression by amino acids and peptides (Hube et al., 1994). Only SAP7 was dramatically derepressed in tup1/tup1 cells (Table 3), indicating that CaTup1 is not required for the repression of other SAP genes. SAP4–6 are closely related, and are generally considered to be co-regulated (Hube et al., 1994; 1997). However, SAP5 expression was depressed in nrg1/nrg1 cells, whereas SAP4 and SAP6 were not, indicating that SAP4–6 may be differentially regulated under certain circumstances.
Table 3. Regulation of ALS and SAP gene families by Mig1, Nrg1 and Tup1.
a. Fold regulation (Mmut/Mwt) for each gene (in arbitrary units). na, not present on the array; ND, expression not detectable.
Highest expression level observed in any of the strains CAF2–1, BCA2–10, LOZ124 or MMC3 (in arbitrary units).
519 ± 89
547 ± 73
93.1 ± 53
32.2 ± 1.0
51.4 ± 1.0
3.7 ± 1.9
1.5 ± 0.9
1.3 ± 1.0
3.2 ± 0.2
4.2 ± 1.2
2.4 ± 1.4
21.3 ± 8.2
1.1 ± 0.6
29.8 ± 10.0
Three members of the ALS family were not present on the array because they were discovered after our gene annotation: ALS4,6,7 (Hoyer et al., 1998b; Hoyer and Hecht, 2000; Hoyer, 2001). (The gene names do not reflect the order in which they were discovered). Under the conditions analysed, ALS1 and ALS2 were expressed at higher levels than other ALS family members (Table 3). These genes were significantly derepressed in tup1/tup1 cells, but not in mig1/mig1 and nrg1/nrg1 cells. In contrast, the hypha-specific members of this family, ASL3 and ALS8, were regulated both by CaNrg1 and CaTup1. ALS5 expression was not significantly affected by the mutations examined. Therefore, CaTup1 and CaNrg1 play major and differential roles in the regulation of ALS gene family members.
This transcript profiling data set for C. albicans supports current models concerning the mode of action of CaMig1, CaNrg1 and CaTup1. These models suggest that CaMig1 and CaNrg1 target the global repressor CaTup1 to distinct sets of C. albicans genes to repress their transcription (Fig. 7, shaded area). The models are based firmly on the relatively well defined mode of action of ScTup1 in S. cerevisiae (reviewed by Smith and Johnson, 2000). Once the ScTup1-ScSsn6 complex has been targeted to a promoter, it is thought to impose transcriptional repression through a combination of mechanisms (Smith and Johnson, 2000). ScTup1-ScSsn6 can interact directly with the transcriptional apparatus to repress transcription. Alternatively, ScTup1-ScSsn6 can interact directly with histone proteins to organize nucleosomes over the promoter, thereby inhibiting the assembly of the transcriptional apparatus. ScTup1-ScSsn6 might also quench the activity of transcriptional activators. A homologue of ScSsn6 has been identified in C. albicans (Smith and Johnson, 2000; S. Garcia and A. J. P. Brown, unpublished). Therefore, CaTup1 might act in similar ways to repress transcription in C. albicans.
Several observations support this model. Firstly, CaMig1 and CaNrg1 appear to be CaTup1 targeting proteins. CaMIG1 and CaNRG1 are the closest C. albicans homologues of ScMIG1 and ScNRG1 respectively (Zaragoza et al., 2000; Murad et al., 2001). Furthermore, CaMig1 and CaNrg1 are both DNA binding proteins with similar sequence specificities to ScMig1 and ScNrg1, respectively, and CaMIG1 is has been shown to be a functional homologue of ScMIG1 (Zaragoza et al., 2000; Murad et al., 2001). Secondly, CaMig1 and CaNrg1 act primarily as repressors in C. albicans. Transcript profiling revealed that many more C. albicans genes were derepressed in response to mig1/mig1 and nrg1/nrg1 mutations than were repressed (Fig. 3). Thirdly, transcript profiling also revealed that CaMig1-CaTup1 regulate different genes from CaNrg1-CaTup1 (Figs 5 and 6, Table 1). Most significantly, CaMig1 and CaNrg1 sites were observed in those promoters that were most tightly controlled by these regulators (Table 1; Murad et al., 2001). This strongly suggests that CaMig1 and CaNrg1 interact directly with these promoters to regulate their expression (Fig. 7).
Some CaTup1-regulated genes were not derepressed in the mig1/mig1 or nrg1/nrg1 mutants (Fig. 5), indicating that CaTup1 regulates some C. albicans genes in a CaMig1- and CaNrg1-independent fashion. This was to be expected because a number of additional ScTup1 targeting proteins have been identified in S. cerevisiae, such as a1/α2, α2/Mcm1, Crt1, Sko1 and Rox1 (Smith and Johnson, 2000). Possible homologues of such genes in C. albicans might target CaTup1 to genes that are not regulated by CaMig1 or CaNrg1 (Fig. 7). Indeed, a Rox1 homologue, Rfg1, was identified recently in C. albicans (Kadosh and Johnson, 2001).
Some C. albicans genes were derepressed in mig1/mig1 or nrg1/nrg1 cells, but not in the tup1/tup1 mutant (Fig. 5). This implies that CaMig1 and CaNrg1 regulate some C. albicans genes in a CaTup1-independent fashion (Fig. 7). At least two possible mechanisms could account for this CaTup1-independent mode of repression. Firstly, CaMig1 and CaNrg1 might employ an alternative, as yet unidentified factor to impose repression at some C. albicans promoters. Secondly, the juxtaposition of activator and operator sites within some C. albicans promoters might allow CaMig1 (or CaNrg1) binding to interfere directly with the binding of an activator(s), even in the absence of CaTup1. These explanations are not mutually exclusive.
Initially, we predicted that CaMig1 and CaNrg1 would regulate different sets of C. albicans genes. As expected, we identified genes that were regulated by either CaMig1 or CaNrg1 (Fig. 5). However, some other C. albicans genes were depressed in mig1/mig1, nrg1/nrg1 and tup1/tup1 cells (Fig. 5), indicating that these genes might be regulated both by CaMig1-CaTup1 and CaNrg1-CaTup1 complexes (Fig. 7).
Detailed studies of ScTup1 function in S. cerevisiae have focused mainly on its role as a transcriptional repressor (Smith and Johnson, 2000). However, a significant number of C. albicans genes appear to be activated by CaTup1 (Fig. 3). The raw transcript profiling data for a S. cerevisiae tup1 mutant (DeRisi et al., 1997) reveal that some genes in budding yeast are subject to positive regulation by ScTup1. It is conceivable that Tup1 might activate the expression of some genes by repressing the activity of a repressor. Alternatively, Tup1 might be able to activate transcription in a context-dependent fashion. The ScTup1-ScSsn6 complex can operate as a transcriptional co-activator on at least one S. cerevisiae promoter under certain circumstances. ScTup1-ScSsn6 was shown to activate CIT2 transcription in response to mitochondrial dysfunction (Conlan et al., 1999). Also, Rap1 and Ume6 have been shown to act as transcriptional repressors or activators in a context-dependent fashion in budding yeast (Hardy et al., 1992; Washburn and Esposito, 2001).
An analysis of the functions of CaTup1-, CaMig1- and CaNrg1-regulated genes provided important clues about the global roles of these regulators. Both CaMig1 and CaNrg1 contribute to the control of metabolic functions, but CaMig1 seems to specialize in the regulation of genes involved in the uptake and utilization of carbon sources and energy production (Fig. 6; Table 1). Consistent with this, mig1/mig1 cells display reduced rates of oxygen utilization, and derepressed expression of the gluconeogenic PEP carboxykinase gene (CaPCK1;Murad et al., 2001). CaNrg1- and CaTup1-regulated genes were enriched in cell rescue functions (Fig. 6), and consistent with this, nrg1/nrg1 and tup1/tup1 cells displayed sensitivity to an oxidative or ethanol stress (Table 2). These data reinforce the idea that CaMig1 and CaTup1 play roles in carbon catabolite repression in C. albicans, whereas CaNrg1 and CaTup1 contribute to some stress responses. This is consistent with the roles of their homologues in S. cerevisiae (Treitel and Carlson, 1995; Marquez et al., 1998; Gancedo, 1998).
Most significantly, the transcript profiling data confirmed that CaNrg1 also regulates genes involved in hyphal development. Four out of the five known hypha-specific genes (ALS3, ALS8, ECE1, HWP1) are strongly regulated by CaNrg1 and CaTup1 (Fig. 5). This corroborates previous reports that these factors repress hypha-specific genes in C. albicans (Sharkey et al., 1999; Braun and Johnson, 2000; Murad et al., 2001), and lends further credence to the quality of the transcript profiling data. Furthermore, these data are consistent with the findings that tup1/tup1 and nrg1/nrg1 mutations derepress filamentous growth in C. albicans (Braun and Johnson, 1997; Murad et al., 2001). Therefore, CaNrg1 and CaTup1 play important roles in the regulation of yeast-hypha morphogenesis (Fig. 7). Unlike C. albicans nrg1/nrg1 mutants (Murad et al., 2001), S. cerevisiae nrg1 mutants do not display an obvious morphological phenotype (M. Straffon and A. J. P. Brown, unpublished). Therefore, the biochemical activities of Nrg1, namely its ability to bind C3T-related sequences and to interact with a Tup1 complex, appear to have been conserved, but the cellular roles of Nrg1 in S. cerevisiae and C. albicans appear to have diverged. A similar picture is emerging for the homologues Rfg1 and Rox1, which seem to regulate different cellular processes in C. albicans and S. cerevisiae respectively (Kadosh and Johnson, 2001).
The transcript profiling data also suggest that CaNrg1 and CaTup1 regulate other virulence attributes, in addition to morphogenesis (Fig. 7). Iron assimilation is known to promote the virulence of C. albicans (Ramanan and Wang, 2000), and genes involved in this process (CFL2, FTR1) are strongly regulated by CaNrg1 and CaTup1 (Fig. 5). Also, these factors differentially regulate members of the ALS and SAP genes families (Table 3; Fig. 5), which promote fungal and enhance virulence (Hube et al., 1997; Sanglard et al., 1997; Hoyer, 2001; P. Leng and A. J. P. Brown, unpublished). We have shown recently that the inactivation of CaNrg1 or CaTup1 attenuates the virulence of C. albicans (Murad et al., 2001). Therefore, both transcript profiling and phenotypic analyses indicate that CaNrg1 and CaTup1 make important contributions to the pathogenicity of this fungus.
To date, only a single target of ScNrg1-mediated repression, STA1, has been defined in S. cerevisiae (Park et al., 1999), and as a result, there is no global perspective on the cellular role(s) of ScNrg1 in budding yeast. This study illustrates, once again, how transcript profiling can be used to rapidly advance our understanding of the biological roles of specific transcriptional regulators. Following genome closure and the production of whole genome arrays, transcript profiling will prove a particularly powerful tool for the analysis of fungal behaviour during the establishment and progression of C. albicans infections.
C. albicans gene annotation
A total of 6213 protein sequences predicted from the S. cerevisiae genome sequence were downloaded from MIPS (http://www.mips.biochem.mpg.de/proj/yeast; 25 April 1995). C. albicans sequences were obtained from two different sources: 380 non-redundant entries of C. albicans ORFs were retrieved from GenBank (http://ncbi.nlm.nih.gov; 9 July 1999); and Assembly 3 was obtained from the Stanford C. albicans sequencing project (http://www-sequence.stanford.edu/group/candida/ 22 April 1999). Assembly 3 contained 1919 contigs of at least 2 kb, covering 12 301 999 bp, or around 80%, of the C. albicans genome, assuming a haploid genome size of 15.5 Mb. The largest contigs (Contig3–3189 to Contig3–3718), representing 6 992 176 bp, were annotated using the following two approaches. Firstly, ORFs were identified using the graphical analysis tool orffinder (http://www3.ncbi.nlm.nih.gov/gorf/gorf.html). Secondly, each contig was searched for segments matching known C. albicans or S. cerevisiae genes. Assembly 3 was compared with C. albicans entries using blastn, and to S. cerevisiae ORFs using blastx using the alternative yeast nuclear code (Ohama et al., 1993). All blast searches were automatically launched and displayed using the scripts blastallgenomes and readblast (Tekaia et al., 2000), which provided a working annotation table listing the ORF name, its length, the percentage identity/similarity/gap and the position of the beginning and end of the match in the contig and in the ORF. paintblast was used to compare orffinder and blast outputs (http://hercules.tigem.it/Biomodules/PaintBlast.pm), which facilitated identification of possible frameshifts or introns. Several classes of C. albicans ORFs were annotated: known C. albicans genes, homologues of S. cerevisiae genes and additional ORFs larger than 150 codons. Overlapping ORFs on the same strand were assumed to result from sequencing errors and considered as a single ORF. The longest ORF was retained when two ORFs overlapped on opposite strands. 3′-truncated ORFs that lay at the end of contigs were discarded because 3′-ends of ORFs were to be arrayed. A total of 3313 putative C. albicans ORFs were identified, and 2016 of these were selected for the construction of gene arrays.
C. albicans array construction
PCR products were arrayed using published procedures (Richmond et al., 1999). PCR primers of 18–22 bases in length were designed using primer3 software (http://www-genome.wi.mit.edu/genome_software/other/primer3.html) to amplify a 3′-region of 300–400 bp from each ORF. These oligonucleotides were synthesized with a 5′-tag (5′-oligonucleotide, 5′-CGACGCCCGCTGATA: 3′-oligonucleotide, 5′-GTCCGGGAGCCATC′) to facilitate subsequent re-amplification of the PCR products. Using these oligonucleotides, the ORFs were PCR-amplified from the C. albicans SC5314 genome. The purity and length of all PCR products were checked by agarose gel electrophoresis. A total of 2002 PCR products, which satisfied our quality controls, were spotted in duplicate onto nylon membranes using a BioGrid System (BioRobotics) (Fig. 1). Candida albicans genomic DNA, E. coli ORFs and S. cerevisiae ORFs were included on the membranes as controls. Detailed information on the array is available at http://http://www.pasteur.fr/recherche/unites/RIF/transcriptdata. The sequences of the PCR primers for each ORF are available upon request.
The following isogenic C. albicans strains were analysed in this study: SC5314 (wild-type clinical isolate: Gillum et al., 1984); CAF2–1 (URA3/ura3::λimm434: Fonzi and Irwin, 1993); BCA2–10 (ura3::λimm434/ura3::λimm434, tup1::hisG/tup1::hisG-URA3-hisG;Braun and Johnson, 1997); LOZ124 (ura3::λimm434/ura3::λimm434, his1::hisG/his1::hisG, mig1::HIS1/mig1::hisG-URA3-hisG;Zaragoza et al., 2000); MMC3 (ura3::λimm434/ura3::λimm434, nrg1::hisG/nrg1::hisG-URA3-hisG). MMC3 was constructed using the methods of Fonzi and Irwin (1993). The nrg1/nrg1 mutation, which changed the start codon from ATG to GTG and deleted 438 bp of the coding region, was confirmed by PCR, Southern blotting and Northern analysis (Murad et al., 2001).
Candida albicans strains were grown at 30°C in YPD (Sherman, 1991) to mid-exponential phase (OD600 = 0.8), and then subcultured into fresh YPD at 30°C and re-grown to an OD600 of 0.8. The cells were then harvested, frozen rapidly in liquid N2, sheared mechanically using a microdismembrator (Braun Melsungen) and RNA prepared by extraction with Trizol ReagentTM (GibcoBRL), as described previously (Hauser et al., 1998). Hybridization probes were made by preparing [33P]-first strand cDNA using a (dT)15 primer (Hauser et al., 1998). cDNA probes were denatured by incubation in 0.2 M NaOH at 65°C, and hybridized to the C. albicans arrays for 24 h at 65°C in 0.5 M NaPO4, 7% SDS, 0.01 M ETDA, pH 7.2 (Church and Gilbert, 1984). Hybridization signals detected using a FUJI FLA-3000 phosphorimager (Raytek). Probes were stripped from membranes using six washes with 0.1% SDS, 5 mM NaPO4, pH 7.2 at 100°C. Duplicate membranes were probed four times without detectable loss of signal.
Phosphorimages were analysed using arrayvision software (Amersham). To control for differences in the specific activity of cDNA probes, quadruplicate signals for each ORF (SignalX) were normalized against those for CAF2–1 using the total signal (ΣSignal1−2002) for each hybridization. Means (M) and standard deviations were then calculated, and genes that showed no significant expression were excluded from further analysis (151 out of the 2002 genes analysed; 7.5%). Correlation coefficients for duplicate hybridizations were 0.938 for CAF2–1, 0.962 for BCA2–10, 0.947 for LOZ124 and 0.964 for MMC3. The Fold Repression (R) for each ORF was calculated using R = Mmutant/Mwt. The data are available at http://www.pasteur.fr/recherche/unites/RIF/transcriptdata/. The analysis of promoter regions was performed using Regulatory Sequence Analysis Tools (http://www.ucmb.ulb.ac.be/bioinformatics/rsa-tools/:van Helden et al., 2000).
To assess sensitivity to different stresses, doubling times for C. albicans CAF-2 were compared with those for strains BCA2–10 and MMC3 during growth at 30°C in YPD containing no additive, 10 mM 3-aminotriazole, 0.7 M NaCl, 4.4 mM H2O2, or 5% ethanol. Experiments were performed in triplicate, and errors were less than 10%.
To compare oxygen utilization rates, the C. albicans strains CAF-2, LOZ124 and MMC3 were grown to mid-exponential growth phase in YPD at 30°C. Cells were harvested, washed and resuspended in 50 mM glucose, 36 mM KH2PO4, pH 7.4, at OD620 ≈ 0.5. Oxygen consumption was measured at 30°C using DW1 Clark-type O2 electrode using the manufacturer's instructions (Hansatech Ltd, King’s Lynn, UK).
We are grateful to L. Frangeul for help with C. albicans gene annotation, to Andrée Lépingle for help with sequencing, to R. Robert for the C. albicans cDNA library and to Mark Ramsdale for performing the oxygen electrode assays. We also thank B. Braun, A. Johnson, O. Zaragoza and C. Gancedo for providing C. albicans strains, and S. Budge for excellent technical assistance. We are grateful to N. Hauser and J. Hoheisel for providing access to arrayvision software. We also thank the Stanford DNA Sequencing and Technology Center for access to its C. albicans genome sequence data, which was generated with the support of the NIDR and the Burroughs Wellcome Fund. AMAM was supported by the Malaysian Government (UKM(PER)7371). HT was supported by the European Commission (EUROFAN, BIO4-CT97–2294) and the UK Biotechnology and Biological Sciences Research Council (1/G11780). AB was supported by The Wellcome Trust (055015, 015108). CE and CG were supported by the French Ministère de la Recherche (Réseau Infections Fongiques, P.R.F.M.M.I.P.). AB, CE and CG were also supported by the European Commission (QLRT-1999–30795).