Converting the complete genome sequence of Candida albicans into meaningful biological information will require comprehensive screens for identifying functional classes of genes. Most systems described so far are not applicable to C. albicans because of its difficulty with mating, its diploid nature, and the lack of functional random insertional mutagenesis methods. We examined artificial gene suppression as a means to identify gene products critical for growth of this pathogen; these represent new antifungal drug targets. To achieve gene suppression we combined antisense RNA inhibition and promoter interference. After cloning antisense complementary DNA (cDNA) fragments under control of an inducible GAL1 promoter, we transferred the resulting libraries to C. albicans. Over 2,000 transformant colonies were screened for a promoter-induced diminished-growth phenotype. After recovery of the plasmids, sequence determination of their inserts revealed the messenger RNA (mRNA) they inhibited or the gene they disrupted. Eighty-six genes critical for growth were identified, 45 with unknown function. When used in high-throughput screening for antifungals, the crippled C. albicans strains generated in this study showed enhanced sensitivity to specific drugs.

Human fungal infections have dramatically increased over the last 15–20 years and have become a significant cause of disease and mortality¹. They are frequently acquired by immunocompromised patients (such as those receiving chemotherapy, undergoing organ transplants, or infected by HIV) and are commonly diagnosed after invasive medical surgery². Currently used antifungal drugs (polyenes such as amphotericin B, triazoles such as itraconazole and fluconazole, and imidazoles such as miconazole) are limited in their use because of their activity spectrum, toxicity, and side effects^{3, 4}. The usefulness of the azoles has significantly diminished in recent years as a result of the increasing incidence of resistance, a complex phenomenon that involves several molecular mechanisms and may be difficult to avoid for this reason. The increase in infections, combined with the reduced efficacy of the currently available drugs, highlights the need for new antifungal drugs with distinct modes of action. The development of such drugs would greatly improve the quality of life and life expectancy of infected patients.

In this study we specifically outline an approach to identify new antifungal drug targets of C. albicans, the major pathogen causing human fungal infections. Although a vast amount of sequence information from the C. albicans genome is now available in both public (http://www-sequence.stanford.edu/group/candida/ ; 79% complete) and private (Incyte Genomics Inc.; 89% complete) databases, "en masse" disruption techniques used to elucidate gene function in other organisms (e.g., the "mass murder" and "genetic footprinting" technology in Saccharomyces cerevisiae^{5, 6}) are not applicable to C. albicans, for several reasons. First, although mating can be induced in C. albicans by creating alterations at the mating type locus^{7, 8}, it is under normal circumstances not capable of mating, so that genetic crosses remain difficult. Second, no convenient method is available for random insertional mutagenesis because of a paucity of functional transposable elements⁹. Third, due to the organism's diploid nature, multiple consecutive steps of gene inactivation are required for making genetic knockouts, and this can only be done on a gene-by-gene basis¹⁰. Whereas existing genetic disruption techniques^11-13 are highly effective for analyzing individual genes, applying them to all of the more than 7,800 predicted C. albicans open reading frames (ORFs) would be a tremendous task¹⁴.

Here we describe an approach to alter gene function on a genome-wide scale in C. albicans genes by using antisense RNA and promoter interference. Repression of gene expression by promoter interference (i.e., gene of interest flanked by two convergent promoters) can occur by physical collision of elongating RNA polymerases on opposite strands¹⁵. Under these circumstances, complementary antisense RNA may be a by-product of the primary transcriptional-level interaction. Inhibition by antisense RNA has not been used extensively in yeast^{16, 17} partly because of the availability of alternative tools for deletion or silencing of genes. We describe a new integrative vector allowing inducible transcription of antisense RNA from a cDNA insert in C. albicans. An antisense cDNA library was constructed using this vector and introduced into C. albicans. Depending on the site of integration in the C. albicans genome, the introduction of such a library plasmid and subsequent induction of the promoter will lead to inhibition of expression of the gene whose cDNA is represented in the library plasmid—the latter either by antisense RNA alone or by a combination of antisense RNA and promoter interference. This approach thus specifically relies on lowering the level of specific C. albicans mRNAs by either of the above mechanisms, thereby decreasing the expression level of the corresponding protein. If this C. albicans protein is critical for growth, the cell will grow more slowly or die. We used this approach in a genome-wide search for new C. albicans genes critical for growth of the pathogen. Among over 2,000 transformant colonies screened, 10% showed a clear growth defect.

Screening for growth-affected transformants. Libraries were introduced into C. albicans strain CAI-4 (ref. 19), and transformants were screened for reduced growth upon activation of the GAL1 promoter in the presence of lithium acetate. Lithium acetate prolongs the G1 phase of the cell cycle, during which antisense is presumed to act most strongly²⁰. More than 2,000 transformants were screened by parallel measurement of growth in noninducing and inducing media. Growth curves of transformants impaired in growth are shown in Figure 3. Candida albicans uses galactose (inducer) rather inefficiently as a carbon source (i.e., the wild-type strain shows reduced growth in antisense induction medium independent of any existing antisense inhibitory effect). To facilitate selection of growth-impaired transformants, maltose was added to both inducing and noninducing media in a later stage of the screening process. Growth of the parental strain CAI-4 was identical in both of these media (data not shown), and the presence of maltose did not significantly influence the induction ratio of the GAL1 promoter (Fig. 4). Approximately 10% (198) of the transformants showed a growth defect and were selected for further analysis. Forty-three percent grew slowly only in inducing medium (consistent with putative promoter interference or antisense effect). Screening with the gDNA library was abandoned because of repetitive isolation of identical DNA inserts (autonomously replicating sequences (ARSs) and ribosomal RNAs (rRNAs); see below) in those transformants showing a growth defect.

Isolation of integrated antisense library clones from the C. albicans genome. To identify those genes whose inhibition causes a growth defect, we needed to investigate which library clones (cDNA inserts) had integrated at which location in the genome of those transformants showing reduced growth. cDNA inserts from the integrated antisense library were isolated from the disruptants by PCR. For integration at the GAL1 promoter (Fig 2A; 11% of all transformants), PCR with 5pGAL1PNiST.PCR and 3pGAL1PNiST.PCR primers amplified the cDNA insert. Because these primers are divergently oriented if integration occurs at the cDNA insert (Fig. 2B), we used inverse PCR (see Experimental Protocol for details).

Some transformants yielded PCR products of different lengths implying multiple integrations. In that case, each of the disrupted genes had to be inactivated separately to find the one causing the growth defect.

From the selected transformants obtained with the gDNA library, ARS2 and 18S or 25 rRNA genes were repeatedly isolated (43, 13, and 8 times, respectively; Table 1), so that we abandoned this approach. The Candida genome contains multiple copies of both ARS2 and rRNA sequences. This would account for their over-representation in the gDNA library but not for their growth-inhibitory effect. However, extrachromosomal rDNA circles (ERCs) accumulate in old yeast cells and actually cause ageing²⁴. Whether there is a relation between this observation and ours is subject to further study.

Many of the genes we identified using the cDNA library are known to be essential in S. cerevisiae or in other organisms (e.g., ribosomal proteins, RPS7, RPL37, RPL27, RPS21, RPL16; translation elongation factors, EFB1, TEF3, TEF4, TUF1; others such as ABP1, RHO1, RNR1, YAE1, TRA1, MEG1; Table 1). Genes involved in carbon source metabolism and nutrient uptake (e.g., a galactose permease, HXT6) were identified as well.

We chose as examples one transformant with an inducible (clone 36; see growth curves in Fig. 3, wells 02 and 102, respectively) and one with a noninducible (clone 38; see Fig. 3 wells 03 and 103, respectively) growth defect for further discussion.

In clone 36, recombination occurred at the cDNA insert as shown by amplification of a 600 base pair gene fragment by inverse PCR. The sequence of this fragment was 74% identical to the S. cerevisiae S-adenosyl methionine synthetase 2 (SAM2) gene. Growth on 5-fluoroorotic acid (i.e., excision of the integrated plasmid from the genome by homologous recombination between the duplicated genes; Fig. 2B) completely restored the wild-type growth phenotype (data not shown), supporting the specificity of the observed growth defect. Northern blot analysis revealed a 0.9 kilobase SAM2 antisense transcript in clone 36 that was absent in the wild-type strain (Fig. 5A). The presence of low amounts of antisense transcript in clone 36 under noninducing conditions is in agreement with the observed leakiness of the GAL1 promoter. Northern blot analysis (Fig. 5B) further revealed a reduced amount of a 1.3 kilobase SAM2 sense transcript (SAM2 mRNA) in clone 36 compared to the wild-type strain when grown in antisense-inducing medium. This is to be expected when sense and antisense transcripts interact to form RNA hybrids (which might subsequently be degraded). To verify the length of SAM2 mRNA, a clone containing the complete ORF (1,155 base pairs) of the SAM2 gene, including 5'- and 3'-flanking regions, was isolated by hybridization screening of a C. albicans gDNA library. Based on a putative TATA box at -27 base pairs and a T-rich (>10 base pairs) region (element described in yeast as necessary for transcript release²⁵) downstream of the ORF, a total transcript length of 1.3 kilobases could be predicted, which is in agreement with what we found. Both the presence of SAM2 antisense RNA and the reduced SAM2 mRNA level upon promoter activation in clone 36 clearly suggest inhibition of SAM2 expression by interference at RNA level.

Clone 38 showed a noninducible growth defect, which is expected when the integration event by itself leads to inhibition of gene expression (e.g., if mutations were present in the original library clone, the protein encoded by the gene after homologous recombination would be nonfunctional; gene suppression would then occur independently of GAL1 promoter activation). PCR analysis revealed the presence of a cDNA library insert (340 base pairs) encoding ribonucleotide reductase 1 (RNR1). Northern blot analysis (Fig. 5C) showed a reduced level of RNR1 mRNA in clone 38 compared to wild-type strain, which was confirmed by real-time quantitative PCR using an RNR1 fluorogenic probe (Fig. 5D). This clearly suggested inhibition of RNR1 expression by interference at the RNA level. To further support the specificity of the observed growth defect in clone 38, revertants were generated on 5-fluoroorotic acid (excision of the integrated plasmid from the genome by homologous recombination between the duplicated GAL1 promoter regions; see Fig. 2A) and showed restoration of the wild-type growth phenotype (data not shown).

Many of the genes that were identified using the screening approach outlined here have an essential function (Table 1), suggesting that the approach succeeds in the identification of genes critical for growth of the pathogen. To support this conclusion, heterozygous knockouts were made in six randomly chosen genes (TEF3, TUF1, RPL27, RHO1, FAL1, and one hypothetical protein (HYP); all identified in the screening outlined here) using the URA-blaster disruption method¹¹. Four disruptants (rho1/RHO1, tuf1/TUF1, rpl27/RPL27 and hyp/HYP) showed a clear growth defect indicating that a gene critical for growth had been targeted. Such a growth defect is seldom observed in heterozygous knockouts of genes randomly chosen from the C. albicans sequence database (unpublished results). The tef3/TEF3 heterozygote strain showed a very clear growth defect when grown on solid growth medium; however, this effect was less apparent in liquid culture. The fal1/FAL1 (ref. 13) heterozygote strain, however, did not show a pronounced growth defect. A possible explanation is that a more pronounced inhibitory effect can be expected with antisense RNA (both copies are targeted) as compared to single-copy inactivation. Therefore, experiments to regulate expression of the second allele of FAL1 (by promoter replacement) are ongoing.

Use of crippled C. albicans strains in drug screening. The genetically crippled C. albicans strains that were generated in this study were used in high-throughput screening for antifungal drugs. Such a high-throughput screening assay typically involves measurement of growth of a genetically crippled strain relative to a wild-type strain in the presence of various compounds, and is based on observations in bacteria and in yeast suggesting that underexpression of any component of a process leads to increased sensitivity to an inhibitor of a relevant step in that process^26-28. Lowering the dosage of a specific gene in C. albicans results in a heterozygote that is sensitized to a drug that acts on the product of the targeted gene. This method thus provides a more sensitive means to identify test compounds with antifungal activity and gives an indication of the site or pathway at which the compounds exert their effect.

Previously Jensen-Pergakes et al.²⁹ showed that a C. albicans sterol methyltransferase (ERG6) mutant was hypersusceptible to a number of sterol synthesis and metabolic inhibitors, including terbinafine, tridemorph, fenpropiomorph, fluphenazine, cycloheximide, cerulenin, and brefeldin A. Similarly, an ERG3 mutant was found to be hypersensitive to certain azole antifungals (Frank Odds and Bart Van Den Hazel, personal communication).

Several crippled C. albicans strains generated in this study (a representative data set is shown in Table 2) showed enhanced sensitivity to specific drugs. Besides strongly inhibiting one crippled strain, some compounds also caused a weak to modest inhibition of other crippled strains. Apart from random variation in growth, this could be due to nonspecific inhibitory effects or some cross-reactivity (virtually no pharmacological compound is 100% specific, especially at high concentrations). Analogs of the compounds we identified are now being synthesized and screened for activity against wild-type Candida strains.

In conclusion, the strategy presented in this study demonstrates a successful approach to alter gene function on a genome-wide scale in an important human pathogenic fungus. Genes critical for growth of C. albicans were successfully identified, and the specificity of the observed growth defects was supported by both rescue experiments and the creation of heterozygous knockouts. This approach, which requires no prior sequence information, provides a tool for functional analysis in some organisms for which existing techniques for "en masse" gene disruption are not applicable. Although the outlined strategy may be adaptable to any diploid organism, imperfect fungi might prove ideal candidates. In-depth investigation (including exploration in animal models) of the targets and drugs outlined in this study is now ongoing.

In vitro drug susceptibility testing. Yeast cultures were grown at 30°C while shaking at 250 r.p.m. until a final OD of 0.2 (0.1) was reached. At this point, 200 l of yeast suspension were added to MW96 plates containing R-compounds in a total volume of 50 l, and plates were incubated (static) at 30°C for 48 h. Growth in positive control (compound-free) and test wells was determined using OD₆₂₀ turbidity readings. (For the purpose of duplicating experiments described in this paper, compounds are available from the Department of Collaboration and Technology Transfer, Janssen Pharmaceutica, Beerse, Belgium.)

Isolation of gDNA or cDNA inserts. gDNA was isolated from transformants using the Nucleon MI Yeast kit (Amersham, Little Chalfont, UK), and 20 ng were digested for 3 h with an enzyme that cuts uniquely in the library vector (SacI for the gDNA library; PstI for the cDNA library). Samples were phenol/chloroform extracted, precipitated with NaOAc/ethanol, resuspended in 500 l ligation mixture (1 ligation buffer and T4 DNA ligase; both from Boehringer Mannheim, Indianapolis, IN) and incubated overnight at 16°C. After denaturation (10 min 65°C), purification (phenol/chloroform extraction), and precipitation (NaOAc/ethanol), the pellet was resuspended in 10 l MilliQ (Millipore, Bedford, MA) water. Inverse PCR was done on 1 l of the precipitated ligation reaction using library vector-specific primers (Fig. 1A, B) (3pGAL1PSiST.PCR: 5'-GAG-GGC-GTG-AAT-GTA-AGC-GTG-3' and 5pGAL1PNiST.PCR: 5'-GAG-TTA-TAC-CCT-GCA-GCT-CGA-C-3' for the gDNA library; 3pGAL1PNiST.PCR: 5'-TGA-GCA-GCT-CGC-CGT-CGC-GC-3' and 5pGAL1PNiST.PCR for the cDNA library; all primers from Eurogentec, Seraing, Belgium).

PCR conditions. PCR was done for 30 cycles each consisting of (a) 1 min at 95°C, (b) 1 min at 61°C, (or 57°C for the cDNA library primers), and (c) 3 min at 72°C. In the reaction mixture 2.5 units of Taq polymerase (Boehringer) with TaqStart antibody (Clontech, Palo Alto, CA) (1:1) were used, and the final concentrations were 0.2 M of each primer, 3 mM MgCl₂ (Perkin Elmer Cetus, Foster City, CA) and 200 M dNTPs (Perkin Elmer Cetus). All PCR reactions were performed in a Robocycler (Stratagene, La Jolla, CA). PCR analysis was also performed on gDNA isolated from the transformants using primers 3pGAL1PSiST.PCR and 5pGAL1PNiST.PCR for the gDNA library transformants and using primers oligo23': 5'-TGC-AGC-TCG-ACC-TCG-AGG-3' and oligo25: 5'-GCG-TGA-ATG-TAA-GCG-TGA-C-3' (Thybr. = 53°C) for the cDNA library transformants. Resulting PCR products were purified using the PCR purification kit (Qiagen, Basel, Switzerland).

Sequence determination and analysis. DNA sequencing was performed as described³⁰. Sequence similarity searches against public (EMBL, SWISS-PROT, TrEMBL, and ALCES (Stanford University, University of Minnesota)) and commercial (LifeSeq and PathoSeq (Incyte Genomics Inc., Palo Alto, CA), and GENESEQ (Derwent, London, UK)) sequence databases were performed using BLAST. The 5' untranslated region of the SAM2 gene was analyzed using the "Findpatterns" algorithm (GCG, University of Wisconsin, Madison).

mRNA quantitation. DIG-labeled RNA probes were prepared by in vitro RNA transcription from 1 g RNAse-free, NsiI-linearized template DNA (SAM2part/pDP19) using the Maxi script kit (Ambion, Austin, TX). DIG-labeled DNA probes were prepared by PCR³⁰. Five micrograms of total RNA from each sample was loaded and probed as described³⁰. Normalization was done by hybridization to an ACT1-specific DNA probe. PCR quantitations using specific primers and probes were performed according to the TaqMan procedure^{34, 35} as described³⁰. Fluorogenic probe for RNR1: 5'-TGA-TCT-CAA-AAA-GTG-CTG-GAG-GAA-TCG-GT-3'; forward primer for RNR1: 5'-CGA-CAC-TTT-GAA-ATC-GTG-TGC-T-3'; and reverse primer for RNR1: 5'-GCA-CCG-GTA-GAA-CGA-ATG-TTG-3'. Data were normalized according to ACT1 C_T values.

Isolation of a full-length SAM2 gene. Amplification of 600 bp of SAM2 gene was by PCR from SAM2/pGAL1PNiST-1 (isolated from clone 36) using primers 5- and 3-pGAL1PNiST.PCR. A C. albicans gDNA library (10–23 kb insert size) in YCp50 (S. Dewaele, University of Gent, Belgium) was probed with radiolabeled SAM2 fragment, and positive colonies were selected for plasmid preparation^{36, 37}. Subsequent restriction enzyme analysis identified a 1.1 kb HpaI fragment from clone 36.13.1 covering the entire hybridizing segment.

Construction of gene disruption cassettes for generation of heterozygous mutants. All disruption cassettes for the generation of single-allele knockouts were generated basically as described^{11, 30}. Fragments of TEF3, TUF1, RPL27, RHO1, and HYP (hypothetical protein; clone 80g3) genes were PCR-amplified from CAI-4 gDNA and subcloned into the pCR2.1-TOPO vector (Invitrogen). Intermediate constructs were cleaved so as to release (for each gene) part of the ORF and served as recipients for the 4 kb URA-blaster cassette (released from pMB7 using PvuI (ref. 11) resulting in six gene-specific disruption constructs. Correct integration of the disruption cassettes was confirmed by Southern blot analysis³⁰. The fal1/FAL1 heterozygous knockout was constructed as described earlier¹³.