|Home | About | Journals | Submit | Contact Us | Français|
Candida albicans is the most common human fungal pathogen and causes significant morbidity and mortality worldwide. Nevertheless, the basic principles of C. albicans pathogenesis remain poorly understood. Of central importance to the study of this organism is the ability to generate homozygous knockout mutants and to analyze them in a mammalian model of pathogenesis. C. albicans is diploid, and current strategies for gene deletion typically involve repeated use of the URA3 selectable marker. These procedures are often time-consuming and inefficient. Moreover, URA3 expression levels—which are susceptible to chromosome position effects—can themselves affect virulence, thereby complicating analysis of strains constructed with URA3 as a selectable marker. Here, we describe a set of newly developed reference strains (leu2Δ/leu2Δ, his1Δ/his1Δ; arg4Δ/arg4Δ, his1Δ/his1Δ; and arg4Δ/arg4Δ, leu2Δ/leu2Δ, his1Δ/his1Δ) that exhibit wild-type or nearly wild-type virulence in a mouse model. We also describe new disruption marker cassettes and a fusion PCR protocol that permit rapid and highly efficient generation of homozygous knockout mutations in the new C. albicans strains. We demonstrate these procedures for two well-studied genes, TUP1 and EFG1, as well as a novel gene, RBD1. These tools should permit large-scale genetic analysis of this important human pathogen.
Candida albicans is a major human pathogen that is especially virulent among people with compromised immune systems, such as premature infants, cancer chemotherapy patients, solid-organ transplant recipients, and patients coinfected with human immunodeficiency virus. The direct cost of treating Candida bloodstream infections in the United States. is between $1 and $2 billion annually (34, 51), and despite aggressive therapies, mortality approaches 50% (49, 18, 32, 3). Complicating efforts to treat disseminated candidiasis is a continued ignorance of basic principles of Candida pathogenesis. Unlike many bacterial pathogens, the pathways to virulence of most pathogenic fungi remain largely mysterious. In the case of C. albicans, this situation is largely attributable to its combination of a diploid genome and an incomplete sexual cycle, both of which have hindered forward genetic approaches to the study of virulence.
Several reagents are now available which, in principle, should make it possible to apply powerful genetic approaches to understanding C. albicans pathogenesis. The new tools include a fully sequenced genome (http://www-sequence.stanford.edu/group/candida/), molecular biological techniques, including protocols for gene disruption (17, 52), microarray technology for expression profiling (23, 26, 27, 5, 41), and mammalian animal model systems for both localized and disseminated infection (28, 31). The mouse model of intravenous infection via the tail vein has been particularly useful in assessing C. albicans behavior in disseminated infection and, at least grossly, resembles disseminated infection in the human host. As in the mouse model, clinically significant candidiasis of humans commonly originates intravenously via contaminated catheters in hospitalized patients. As is true for the mouse (40), the brain and kidneys are prime targets of disseminated infection in humans. If the C. albicans genes mediating virulence in the mouse model overlap those responsible for pathogenesis in humans, this model will be key to elucidating new targets for antimicrobial therapy.
In light of the central role of the mouse model in studying C. albicans pathogenesis, a C. albicans genetic system should be compatible with it. For historical reasons, the majority of C. albicans knockout strains have been generated with the selectable auxotrophic marker URA3; however, URA3 expression levels are now known to influence C. albicans virulence (8, 10, 22, 45) as well as morphogenesis and adhesion (4, 10). In some cases, the extent of URA3 expression has been shown to vary with the chromosomal position it occupies (8, 10, 24). Use of URA3 at heterologous loci to mark gene knockouts may therefore confound interpretation of observed virulence defects, and it has been suggested that as many as 30% of published virulence mutants have misattributed virulence defects linked to URA3 to deletions of other genes (8). One proposed solution to this problem is to construct knockout mutants for virulence analysis so that URA3 is restored to a single chromosomal location from which it is consistently well expressed (e.g., the ENO1 or RPS10 locus) (45). Another solution is to use a dominant selectable marker that can be recycled for each round of gene disruption (35). A third solution, which we propose below, is to use alternative auxotrophic markers that do not significantly influence virulence.
A second, more general problem with current C. albicans gene disruption methodology is the absence of reliable and efficient protocols. Because C. albicans is diploid, complete abolition of gene function requires disruption of both alleles of a given target gene. A typical gene disruption protocol (52) uses long PCR primers with 60 to 80 nucleotides of homology to genomic DNA flanking the target genes to amplify the URA3 gene. The URA3-linked disruption fragment is transformed into a Ura− C. albicans strain, and Ura+ transformants are screened for the absence of one allele of the target gene. To recycle the URA3 marker, heterozygous disruption mutants are passaged on medium containing 5-fluoroorotic acid (1, 6), which selects intragenic recombinations that delete the URA3 gene. Exposure of C. albicans to 5-fluoroorotic acid also generates unlinked chromosomal rearrangements and other genetic abnormalities (50a), making this step a potential source of unwanted genetic changes. Finally, the Ura− heterozygous knockout strain is transformed with the same URA3-linked disruption fragment to disrupt the second allele of the target gene. Often the second gene disruption event is inefficient, with the majority of Ura+ transformants resulting from reinsertion of the disruption fragment into the first knockout locus or elsewhere in the genome. In some cases, it has been necessary to screen more than 100 transformants by PCR to obtain a single strain in which the second allele was properly disrupted (our unpublished results).
In this paper, we address these problems through the derivation of new C. albicans strains as well as supporting plasmids and protocols. We describe three newly constructed C. albicans strains that contain homozygous null mutations in HIS1 and LEU2 (strain SN87), HIS1 and ARG4 (strain SN95), or HIS1, LEU2, and ARG4 (strain SN152). These strains have normal karyotypes, as determined by pulsed-field gel electrophoresis, and we show that disruption of HIS1, LEU2, or ARG4 in these strains has minimal effect on C. albicans virulence in the mouse tail vein injection model, permitting use of these genes to mark deletions for virulence analysis. With a fusion PCR technique and homologous complementing HIS1, LEU2, and ARG4 plasmids, we show that genes can be rapidly and efficiently knocked out, providing as specific examples two previously studied genes, EFG1 and TUP1, and one previously uncharacterized gene, RBD1. An important component of the new method is the use of HIS1, LEU2, and ARG4 from non-C. albicans Candida species as selectable markers for the gene disruption steps; the heterologous markers likely suppress undesirable recombination with the chromosomal marker loci. These strains and procedures permit the rapid creation and testing of C. albicans knockout mutants that are suitable for animal studies of C. albicans virulence.
See Table Table22 for the sequences of all PCR primers. Fusion PCR (see below) (50) was performed to create disruption fragments of C. albicans LEU2 and ARG4 coding sequences. For the first round of PCRs (Fig. (Fig.5B),5B), genomic DNA from C. albicans strain CAF2-1 (17) and LEU2 and ARG4 primer pairs 1 and 3 and 4 and 6 were used to amplify the 5′- and 3′-flanking sequences, respectively, for the two genes. URA3 flanked by 200-nucleotide repeats was amplified from pDDB57 (52) with pDDB57 primers 2 and 5. The remainder of the fusion PCR protocol was performed as described below.
The LEU2 and ARG4 disruption fragments were transformed into RM1000#2 by standard methods (39) and plated on medium without uracil. Ura+ transformants were tested by colony PCR for the presence of correct knockout junctions with the following primers: for the 5′ ends, LEU2 Upstream Check or ARG4 Upstream Check plus URA3 Left, and, for the 3′ ends, LEU2 Downstream Check or ARG4 Downstream Check plus URA3 Right. Positives from these diagnostic PCRs were confirmed by colony PCR with primers that flank the disrupted locus (i.e., Upstream and Downstream Check primers); positives in this test had two bands, one of normal genomic size and a second, knockout product that was longer because of the introduction of URA3.
PCR-confirmed heterozygous disruptants were streaked onto 5-fluoroorotic acid-containing medium to recycle the URA3 marker as previously described (6, 1). Ura− strains were transformed with the disruption fragments to knock out the second allele of LEU2 or ARG4. Ura+ transformants were screened as above with the Upstream Check and Downstream Check PCR primers for loss of the shorter PCR product corresponding to the intact gene and persistence of the longer PCR product corresponding to the disrupted alleles. Candidates that passed this test were subjected to a final PCR to exclude the ectopic presence of LEU2 or ARG4 elsewhere in the genome, with primers that amplify sequences within the LEU2 and ARG4 genes, LEU2 open reading frame (ORF) 5′ and LEU2 ORF 3′ or ARG4 ORF 5′ and ARG4 ORF 3′. Strains that were correct by PCR, that had expected growth on appropriate dropout media, and that had a normal karyotype by contour-clamped homogenous electric field gel analysis (below) were designated SN78 (Leu− His− Ura−) and SN76 (Arg− His− Ura−). Note that SN76 is similar to the previously published reference strain BWP17 (53), with the exception that SN76 has a full-length copy of chromosome 5.
Strain SN148 (Arg− Leu− His− Ura−) was derived from SN76 by repetition of the LEU2 gene disruption process, as described above.
URA3 (and the adjacent gene, IRO1) were restored to their normal loci in SN78, SN76, and SN148 by transformation with the 4.9-kb PstI-BglII restriction fragment from pLUBP (33). Transformants were selected on medium without uracil, and integration at the proper locus was confirmed by colony PCR with a primer upstream of the integrated fragment (URA3 Left Flank) and a primer internal to the restored sequence (URA3 Right Flank). The new strains, now heterozygous for wild-type URA3 at the genomic locus, were designated SN87 (Leu− His−), SN95 (Arg− His−), and SN152 (Arg− Leu− His−).
Chromosome plugs were prepared from strains SC5314 (19), RM1000 (29), RM1000#2 (B. B. Magee and P. T. Magee, unpublished), and SN78, SN76, SN87, SN95, SN148, and SN152 as described previously (11). Plugs were introduced into 1% agarose gels, and electrophoresis was performed with a Bio-Rad CHEF-DRIII system. The gels were run in 0.5x Tris-borate-EDTA buffer and maintained at 14°C throughout the procedure. Electrophoresis settings were as follows (adapted from those described on the National Research Council of Canada website for the Whiteway laboratory, http://cbr-rbc.nrc-cnrc.gc.ca/biovis/candida): field angle was 106° throughout; during the first 24 h, voltage was 4.2 V/cm with a switch time of 300 s; for the subsequent 96 h, voltage was 2.4 V/cm with a switch time of 1,200 s.
C. maltosa LEU2 was amplified by PCR with primers C. maltosa LEU2 5′ and C. maltosa LEU2 3′ as well as a template of C. maltosa genomic DNA (prepared from strain 38041, obtained from the American Type Culture Collection). C. dubliniensis HIS1 (cloned before the C. dubliniensis genomic sequence was available) was amplified with primers based on C. albicans genomic sequence, C. albicans HIS1 5′ and C. albicans HIS1 3′, and C. dubliniensis genomic DNA (prepared from strain MYA-577, obtained from the American Type Culture Collection). C. dubliniensis ARG4 was amplified with primers C. dubliniensis ARG4 5′ and C. dubliniensis ARG4 3′ and C. dubliniensis genomic DNA. The C. dubliniensis ARG4 primers were designed with sequences provided by the Wellcome Trust Sanger Institute Pathogen Sequencing Unit, available online at http://www.sanger.ac.uk/Projects/C_dubliniensis/.
The PCR products were ligated via blunt ends to the vector, pCR-BluntII-TOPO, with the Zero Blunt TOPO PCR cloning kit from Invitrogen, creating plasmids pSN40 (C. maltosa LEU2), pSN52 (C. dubliniensis HIS1), and pSN69 (C. dubliniensis ARG4). The PCR inserts pf pSN40 and pSN52 were sequenced, and the nucleotide sequences of C. dubliniensis HIS1 and C. maltosa LEU2 have been deposited in GenBank (accession numbers AY534141 and AY534142, respectively).
For complementation analysis, the C. maltosa LEU2 gene was subcloned into pBES116 (16), a plasmid that interposes the URA3 gene and a polylinker sequence between genomic sequences that flank the ADE2 gene. An EcoRI-EcoRI fragment from pSN40 was blunt ended with T4 DNA polymerase and ligated to EcoRV-cut pBES116. The resulting plasmid was digested with AscI to generate an integrating fragment that targets C. maltosa LEU2 to the C. albicans ADE2 gene locus.
The integrating fragment was transformed into SN78 (Leu− Arg−), and transformants were selected on uracil dropout medium. Ura+ transformants were screened for the ability to grow on leucine dropout medium, as shown for one transformant in Fig. Fig.3B3B (incubated at 30°C for 3 days).
Strains SC5314, RM1000#2, SN100, SN87, SN95, and SN152 were inoculated from overnight cultures 1:30 into fresh YEPD medium and grown for 4 h at 30°C. Logarithmically growing cells were pelleted, washed in sterile normal saline, and counted with a hemacytometer; 8 × 105 cells of each strain were inoculated into the tail veins of 8- to 10-week-old BALB/c mice, seven mice per C. albicans strain. The mice were monitored daily and sacrificed when moribund (defined as hunched posture, minimal motor activity, and weight loss of >15% of starting body weight).
Strains RM1000#2, SN100, SN87, and SN95 were inoculated from saturated overnight cultures into fresh YEPD medium and washed as above. Equal numbers of cells from the four strains were mixed, and 5 × 105 cells of the mixture were inoculated into the tail veins of five BALB/c mice. The mice were monitored daily and sacrificed when moribund. The infected kidneys were recovered, homogenized, diluted, and plated on Sabouraud agar with added ampicillin (50 μg/ml) and gentamicin (15 μg/ml). Recovered CFU were grown for 2 days at 30°C and replica plated to appropriate dropout media to determine the relative numbers of each of the four starting strains.
Statistical analysis of differences between growth curves was performed with the log-rank test.
The fusion PCR strategy (50) is outlined in Fig. Fig.4;4; see Table Table22 for primer sequences. The first round of reactions involves amplification of the flanking sequences of the target gene (with a template of genomic DNA and primers 1 and 3 or 4 and 6, in separate reactions) and the selectable marker (with cloned HIS1, LEU2, or ARG4 and primers 2 and 5). The 5′ tails of primers 2 and 3 are complementary, as are the 5′ tails of primers 4 and 5. In the second or “fusion” round of PCR, all three products of the first round are combined, and a fusion product is amplified with primers 1 and 6.
TUP1, EFG1, and RBD1 gene disruption fragments were created as follows. For the first round of PCRs, 50-μl reactions were assembled with 0.5 μl of KlenTaq LA DNA polymerase (BD Biosciences) in 1x KlenTaq buffer, 0.25 mM deoxynucleoside triphosphates, 0.2 uM each primer, and appropriate template DNA. The template for flanking sequences was CAF2-1 genomic DNA, and the template for C. maltosa LEU2, C. dubliniensis HIS1, and C. dubliniensis ARG4 markers was pSN40, pSN52, and pSN69, respectively. The reactions were incubated as follows: 93°C for 5 min; then 35 cycles of 93°C for 30 s, 45°C for 45 s, and 72°C for 3 min; and finally 72°C for 10 min. The flanking sequence products were purified with the QIAquick PCR purification kit (Qiagen), whereas the marker products were gel purified on a 1% agarose gel prior to purification with the QIAquick gel extraction kit (Qiagen).
For the fusion reactions, 50-μl reactions were assembled as above except that the template consisted of 1 μl of each of the three preceding reaction products, and the primers were primers 1 and 6. Reaction temperatures were as follows: 93°C for 5 min; then 35 cycles of 93°C for 30 s, 50°C for 45 s, and 72°C for 4.5 min; and finally 72°C for 10 min. The fusion PCR product was purified with the QIAquick PCR purification kit prior to transformation into C. albicans strains SN87, SN95, and SN152.
The primers used for diagnosis of the three gene knockouts were, for the 5′ junctions, TUP1 Upstream Check, EFG1 Upstream Check, or RBD1 Upstream Check plus the Left primer for HIS1, LEU2, or ARG4, as appropriate; for the 3′ junctions, TUP1 Downstream Check, EFG1 Downstream Check, or RBD1 Downstream Check plus the Right primer for HIS1, LEU2, or ARG4; and, for trans-ORF PCR, the Upstream and Downstream checks as appropriate for each gene.
Vectors pSN75 and pSN78 were created to facilitate complementation analysis of homozygous disruption mutants by targeting a wild-type version of the appropriate gene to the RPS10 locus, with C. dubliniensis HIS1 or C. maltosa LEU2, respectively, as the selectable marker. One-kilobase regions just upstream and downstream of the RPS10 gene were amplified by PCR with primers RPS10 5′ HindIII and RPS10 5′ SalI, and RPS10 3′ EcoRV and RPS10 3′ SacII, with CAF2-1 genomic DNA as the template. The primers included sites for the named restriction enzymes at their 5′ ends. PCR products were cut with the appropriate restriction enzymes and subcloned into the vector backbone of pAG25 (20). To create pSN75, C. dubliniensis HIS1 was subcloned from pSN52 as a BamHI-NotI (blunt ended at NotI with T4 DNA polymerase) fragment into the BamHI- and EcoRV-cut intermediate plasmid; this placed the selectable marker between the RPS10 upstream and downstream sequences. To create pSN78, C. maltosa LEU2 was subcloned from pSN40 as a SacI-NotI (blunt ended at both sites with T4 DNA polymerase) fragment into the same intermediate plasmid, this time cut with BamHI and EcoRV and blunt-ended with T4 DNA polymerase.
Wild-type RBD1 was amplified by PCR with primers RBD1 5′ XhoI and RBD1 3′ XhoI, cut with XhoI, and subcloned into the XhoI site of pSN75 (between the RPS10 upstream sequence and C. dubliniensis HIS1) to create plasmid pSN84. Wild-type TUP1 was amplified with primers TUP1 5′ SacI and TUP1 3′ SacI, cut with SacI, blunt-ended with T4 DNA polymerase, and subcloned into pSN78 cut with SalI and blunt ended with T4 DNA polymerase to create plasmid pSN83.
The wild-type RBD1 and TUP1 complementation fragments were released from pSN84 and pSN83 with HisI (RBD1) or NotI (TUP1) (both enzymes cut just upstream of the RPS10 upstream flanking sequence) and SacI, which cuts just downstream of the RPS10 downstream flanking sequence, purified with a QIAquick PCR purification kit (Qiagen), and transformed into the rbd1Δ/rbd1Δ and tup1Δ/tup1Δ strains, respectively, selecting on medium lacking leucine, arginine, and histidine.
Growth of the homozygous deletion mutants, addback strains, and CAF2-1 was assessed by plating dilutions of each strain onto Spider medium and incubating the plates at 30°C. Growth of single colonies was assessed after 6, 9, and 12 days.
Sequence data from this article have been deposited with the EMBL and GenBank data libraries under accession numbers AY534141 and AY534142.
RM1000 (ura3Δ/ura3Δ, his1Δ/his1Δ, iro11Δ/iro1Δ) is a strain derived from the clinical isolate SC5314 and is routinely used to make gene disruptions in C. albicans (29) (Table (Table1).1). Recent karyotyping of this strain by B. B. Magee and P. T. Magee and by our laboratory has revealed a large deletion in one copy of chromosome 5 (P. T. Magee, personal communication; A. E. Tsong and A. D. Johnson, unpublished observations). Shown in Fig. Fig.1,1, lanes 1 and 2, are the contour-clamped homogeneous electric field gel results for chromosomes prepared from SC5314 and RM1000, respectively; the increased mobility of one copy of chromosome 5 can easily be observed.
With an independent isolate of RM1000 (called RM1000#2) provided by the Magee laboratory that is wild type for both copies of chromosome 5 (P. T. Magee, personal communication; Fig. Fig.1,1, lane 3), we have derived three new strains for gene disruption and virulence analysis. The genotypes of the strains are presented in Table Table11.
The rationale for strain design was as follows. For reasons outlined in the introduction, we wished to avoid use of URA3 as a marker for gene disruption studies. Therefore we sought alternative auxotrophic markers that would be neutral for virulence in the mouse. HIS1 and LEU2 have previously been reported to have minimal effects on C. albicans virulence (2, 22), and ARG4 was previously untested. With a fusion PCR technique (see below and Materials and Methods) and the vector pDDB57 (encoding a recyclable URA3 gene flanked by 200-nucleotide direct repeats ), we created strains SN78 (leu2Δ/leu2Δ, his1Δ/his1Δ, ura3Δ/ura3Δ), SN76 (arg4Δ/arg4Δ, his1Δ/his1Δ, ura3Δ/ura3Δ), and SN148 (arg4Δ/arg4Δ, leu2Δ/leu2Δ, his1Δ/his1Δ, ura3Δ/ura3Δ). The disruption primers (Table (Table2)2) were designed to precisely delete the open reading frames of the auxotrophic markers while leaving promoter and termination sequences intact, so as to minimize disruptions of neighboring genes. To restore prototrophy for uracil biosynthesis to the three strains, a 4.9-kb BglII-PstI fragment from pLUBP (encoding wild-type URA3 as well as IRO1 ) was introduced into SN78, SN76, and SN148, creating strains SN87 (leu2Δ/leu2Δ, his1Δ/his1Δ, URA3/ura3Δ), SN95 (arg4Δ/arg4Δ, his1Δ/his1Δ, URA3/ura3Δ), and SN152 (arg4Δ/arg4Δ, leu2Δ/leu2Δ, his1Δ/his1Δ, URA3/ura3Δ), respectively. These strains all have one copy of URA3 restored to its normal genetic locus.
At each step of strain construction, integration of the disruption construct at the correct genomic location was verified by PCR of the 5′ and 3′ ends of the fragment to genomic DNA to assess disruption junctions; PCR across the genomic locus to confirm alteration in the size of the disrupted gene; and (for homozygous disruption mutants) PCR with primers internal to the disrupted open reading frame (ORF) to confirm the absence of the ORF at any position in the genome (data not shown). To verify proper restoration of URA3-IRO1 in strains SN87 and SN95, PCR was performed with one primer from within the restored sequence and another from flanking genomic DNA (data not shown).
SN78, SN76, SN87, SN95, SN148, and SN152 behave as predicted on uracil, histidine, leucine, and arginine amino acid dropout plates (Fig. (Fig.2).2). In each case, the strains are unable to grow on medium lacking the amino acid whose synthesis gene was targeted by the disruption but are able to grow normally on YEPD as well as other dropout media. These growth defects independently confirm that the targeted gene disruptions inactivate both alleles of the auxotrophic marker genes.
Chromosomal DNAs prepared from SC5314 RM1000#2 and the six new strains were resolved by contour-clamped homogenous electric field gel electrophoresis, shown in Fig. Fig.11 (lanes 4 through 11). RM1000#2 DNA (Fig. (Fig.1,1, lanes 3 and 5) has two normal-length copies of chromosome 5 (which comigrate with those of SC5314; Fig. Fig.1,1, lanes 1 and 3), as do the six new strains (Fig. (Fig.1,1, lanes 6 through 11). By this assay, all seven numbered chromosomes appear normal in the new strains; however, we cannot rule out small differences in chromosomes that would not be detected by this low-resolution method. Chromosome R—characterized by hundreds of repeats of the rRNA genes—has altered mobility in several strains; such variation in length of chromosome R is commonly observed among clinical and laboratory isolates and is not associated with defective virulence (37).
To assess the effects of sequential deletion of HIS1, LEU2, and ARG4 as well as URA3 on C. albicans virulence in the mouse model, the new strains were compared to each other, to the RM1000#2 parental strain (Ura− His−), and to SN100 (RM1000#2 to which one copy of URA3 has been restored); these results are presented in Fig. Fig.3.3. Except for the indicated differences in auxotrophic markers, these strains are isogenic. To confirm that the derived strains retain wild-type virulence, SN100 was also compared to SC5314, the wild-type clinical isolate that is the ultimate progenitor of the other strains.
Figure Figure3A3A presents a plot of time to illness of seven immunocompetent BALB/c mice infected with 8 × 105 CFU of SN100 (His−) or the Ura− but otherwise isogenic parental strain RM1000#2 (Ura− His−). Whereas all mice infected with SN100 became moribund within 1 week, those infected with RM1000#2 continued to be healthy and to gain weight until the experiment was terminated at 60 days (the results for only the first 20 days are depicted in Fig. Fig.1a).1a). This difference in virulence is highly statistically significant (P = 0.0004) and echoes previous reports of the strong influence of URA3 on virulence (19). In contrast, Fig. Fig.3B3B depicts a comparison between mice infected with SN100 (His−), SN87 (His− Leu−), and SN95 (His− Arg−) under the same conditions. Here, the morbidity curves were quite similar and statistically indistinguishable.
Figure Figure3C3C presents a final comparison between SN100 (His−), the triply auxotrophic strain SN152 (His− Leu− Arg−), and the wild-type clinical isolate SC5314. Here, SN152 appears to have attenuated virulence compared to SN100; for example, the time to sacrifice of half the mice is approximately 8 days for SN152 compared to only 4.5 days for SN100. This difference in virulence is statistically significant (P = 0.0112 by the log-rank test), although of lesser magnitude than the difference between SN100 and RM1000#2. Finally, comparison between the virulence curves of SN100 and SC5314 confirms that SN100 has normal virulence and supports previous reports (2) that auxotrophy for histidine does not significantly affect virulence in the mouse model.
Strains RM1000#2, SN100, SN87, and SN95 were compared in a second set of experiments designed to detect even subtle differences in virulence. Here, the four strains were mixed in equal numbers, and the pooled inoculum was used to infect five mice. Each mouse was sacrificed as it became moribund (after 7 days, on average; data not shown), and viable C. albicans were recovered from mouse kidneys. The representation of each C. albicans strain in the infecting versus recovered pools was determined by growth of cells on amino acid dropout media. As shown in Fig. Fig.3D,3D, the four strains were represented equally in the inoculum; however, in the pool recovered after infection of the mice, RM1000#2 (Ura− His−) was absent, whereas the remaining three strains were represented equally.
SN87, SN95, and SN152 were constructed so that each allele of a target gene may be disrupted with a different genetic marker; SN152 has a third auxotrophic marker available for restoring a wild-type (or mutant) version of the target gene, if desired. Because the genomic disruptions of LEU2 and ARG4 delete only the coding sequence, leaving promoter and terminator sequences intact, we sought complementing genes from other species. The reasoning was that sequence heterogeneity between the promoter and terminator sequences of the complementing and disrupted genes should minimize recombination at the corresponding C. albicans genomic loci.
Using the published genomic sequence of Candida maltosa LEU2 (which retains the same noncanonical CTG codon usage as C. albicans [43, 46]), we synthesized primers to anneal approximately 1 kb upstream and 0.5 kb downstream from the LEU2 coding sequence; the gene was then amplified by PCR with C. maltosa genomic DNA. DNA sequencing revealed some differences from the previously published sequence of C. maltosa LEU2 (presumably due to strain differences) and 61% DNA sequence identity to C. albicans LEU2 over the same region (the new C. maltosa LEU2 sequence has been deposited in GenBank, accession number AY534142).
At the time this work was started, published sequences for HIS1 from organisms with C. albicans CTG codon usage were not available, so we designed primers flanking this gene based on the C. albicans genomic sequence. Using primers that anneal approximately 1 kb upstream and 0.5 kb downstream of C. albicans HIS1, we successfully amplified the cognate gene from C. dubliniensis, the Candida species most closely related to C. albicans (44). Over the region amplified, C. dubliniensis HIS1 is 83% identical to C. albicans HIS1 at the DNA level (deposited as GenBank accession number AY534141). C. dubliniensis ARG4 was cloned with C. dubliniensis-specific primers after the provisional C. dubliniensis genomic sequence was published on the Internet by the Wellcome Trust Sanger Institute Pathogen Sequencing Unit (available online at http://www.sanger.ac.uk/Projects/C_dubliniensis/).
The heterologous HIS1, LEU2, and ARG4 genes were tested for the ability to complement the new C. albicans strains. Each gene fragment was subcloned into pBES116 (16), a vector that directs integration into the Candida ADE2 gene and is linked to a wild-type URA3 gene. Transformants were selected on uracil dropout medium and then tested for growth on histidine, leucine, and arginine dropout plates, as appropriate. Shown in Fig. Fig.2B2B is complementation of SN87 (URA3Δ/ura3Δ, his1Δ/his1Δ, leu2Δ/leu2Δ) by C. maltosa LEU2 on medium without leucine. C. dubliniensis HIS1 similarly complements SN87 on medium without histidine, and C. dubliniensis ARG4 fully complements SN95 (URA3Δ/ura3Δ, his1Δ/his1Δ, arg4Δ/arg4Δ) on medium without arginine (data not shown).
To achieve more efficient gene disruption, we modified a protocol for creating disruption fragments that has previously been used for Saccharomyces cerevisiae and Cryptococcus neoformans (12, 50; O. Liu and H. D. Madhani, personal communication). Our current method (Fig. 4A and B) uses a two-step fusion PCR protocol that results in a disruption fragment with relatively long (>350 nucleotides) regions of homology with the target locus. The first step consists of three PCRs, two to amplify the upstream and downstream regions of the target gene and one to amplify the selectable marker. The internal PCR primers have complementary tails that are used in the second round of PCRs for mutually primed synthesis. The result is a disruption fragment with long regions of homology to the upstream and downstream flanking sequences of the target gene (350 nucleotides in this example).
To test the new disruption technique and strains, primers were designed to disrupt TUP1 and EFG1 in SN87, SN95, and SN152. We chose TUP1 and EFG1 because they have been studied extensively, and the phenotypes of heterozygous and homozygous mutant strains have been well described. To create tup1 mutants of SN87, SN95, and SN152, disruption constructs were prepared with all three heterologous selectable genes, C. maltosa LEU2, C. dubliniensis HIS1, and C. dubliniensis ARG4.
In independent experiments, the first round of gene disruption was carried out with one of the three selectable markers, and the second round was carried out with one of the two remaining markers. After selection of transformants on the appropriate single or double amino acid dropout mediym, gene disruption candidates were screened by PCR for expected 5′ and 3′ junctions as well as the size of the disrupted gene (data not shown). The tup1/tup1 double disruptant was therefore constructed six different ways and, in all cases, the phenotypes of independent mutants were identical. Regardless of the marker used, the first round of gene disruption was approximately 90% efficient; that is, of the transformants that grew on the appropriate amino acid dropout media, 90% carried a precisely disrupted allele, as judged by PCR. By these same criteria, the second round of gene disruption was approximately 70% efficient. The same procedures were repeated for EFG1, and the efficiencies of the first and second gene disruption events were similar to those of TUP1. Again, the phenotypes of newly created efg1/efg1 strains matched those of published efg1 mutants (data not shown). These observations demonstrate that the gene disruption procedures described here are highly efficient and do not require a prescribed order of marker use for constructing double disruptants.
The advantage of using heterologous genetic markers from non-C. albicans yeasts was revealed by the following experiment. If C. albicans ARG4 instead of C. dubliniensis ARG4 was used to mark the first disrupted allele of either TUP1 or EFG1, a significantly lower efficiency (approximately 25%) was observed. Use of C. albicans ARG4 to select for disruption of the second allele was even lower, <1% (data not shown). These results indicate that use of selectable markers from non-C. albicans Candida species greatly improves the efficiency of gene disruption, probably by preventing undesired integration events at the endogenous locus of the auxotrophic marker.
The triply auxotrophic strain SN152 (Arg− His− Leu−) was created to permit disruption of both alleles of a given target gene with two auxotrophic markers; the third marker is available for introduction of another genetic element, such as a wild-type copy of the disrupted gene for complementation analysis. We performed such analysis for the tup1/tup1 mutant. The TUP1/tup1Δ::C. dubliniensis ARG4 heterozygous knockout and tup1Δ::C. dubliniensis HIS1/tup1Δ::C. dubliniensis ARG4 homozygous knockout strains were created in the experiments described above. For complementation analysis, wild-type TUP1 was amplified by PCR and—along with C. maltosa LEU2—subcloned between 1-kb sequences flanking the RPS10 gene; previous studies have established that introduction of exogenous genes to the RPS10 locus is efficient and associated with high levels of gene expression (8). The homozygous disruption strain, tup1Δ::C. dubliniensis HIS1/tup1Δ::C. dubliniensis ARG4, was transformed with the linearized fragment containing wild-type TUP1 and C. maltosa LEU2 and selected on medium without arginine, histidine, or leucine. Transformants were screened by PCR for the presence of TUP1.
Shown in Fig. Fig.55 is the growth of each mutant and the complemented strain versus that of CAF2-1 on Spider medium at 30°C, which weakly induces filamentation of wild-type C. albicans. Under these conditions, the heterozygous and homozygous tup1 knockout strains recapitulate the previously published phenotype of exuberant hyperfilamentation (corresponding to increased numbers of pseudohyphae and seen here as increased wrinkling of the surface of the colony) (9) compared to CAF2-1. A second notable phenotype of tup1 mutants is a progressive increase in the central part of the colony relative to peripheral hyphae, so that peripheral filaments are virtually undetectable in the double mutant. Restoration of one copy of TUP1 partially suppresses both the hyperfilamentation phenotype and the loss of peripheral filaments, so that the complemented strain resembles the heterozygous knockout mutant. This is consistent with previous reports for heterozygous and homozygous tup1 mutants (9) and similar to the haploinsufficiency phenotypes of other C. albicans mutations that affect cell morphology (47).
To determine the utility of the new reagents and methods in constructing C. albicans knockout mutants on a larger scale for use in genetic screens, we constructed 144 mutants in a 96-well format and assessed their colony morphology on Spider medium at 30°C (data not shown). In the initial screen, one mutant, which we have named rbd1/rbd1 (for rhomboid-like), had a phenotype almost opposite that of tup1/tup1, namely, a diminution of the central part of the colony so that almost the entire colony consisted of peripheral filaments. RBD1 (whose two alleles correspond to ORF19.5234 and ORF19.12698 in the Stanford Genome Technology Center sequence) is a previously uncharacterized C. albicans open reading frame with significant homology (BLAST score of 10−32) to a family of genes conserved from archaebacteria to humans, which encode putative intramembrane serine proteases (48). For example, Drosophila melanogaster rhomboid protein P20350 cleaves the transforming growth factor alpha growth factor Spitz, allowing it to activate the epidermal growth factor receptor (38, 48). An S. cerevisiae rhomboid homolog, Rbd1/Pcp1, is localized in mitochondria, where it cleaves proteins localized in the intermembrane space (15, 21, 25).
We constructed a set of independent rbd1/rbd1 homozygous deletion mutants in the SN152 (Arg− Leu− His−) background with C. dubliniensis ARG4 and C. maltosa LEU2 to mark the double disruptants. The phenotype was identical to that described above. Next, we restored one copy of wild-type RBD1 to the RPS10 locus, with C. dubliniensis ARG4 as a selection. Shown in Fig. Fig.55 are the colony morphologies of CAF2-1, RBD1/rbd1, rbd1/rbd1, and rbd1/rbd1 with RBD1 added back at the RPS10 locus after growth on Spider medium at 30°C for 9 days. The phenotypes of the single- and double-disruption mutants are notable for a progressive decrease in area occupied by the central part of the colony, a defect that is partially complemented (back to the intermediate phenotype of the heterozygous disruption mutant) by restoration of one copy of wild-type RBD1. This phenotype was consistent among five independent double knockout mutants and was complemented in all five by restoration of wild-type RBD1 (data not shown).
In this paper, we describe three newly derived strains that are useful for gene disruption and subsequent virulence analysis of C. albicans. Unlike some strains currently used for these purposes, the new strains have a chromosomal complement similar to that of the original clinical isolate, SC5314. They also feature auxotrophic markers, HIS1, LEU2, and ARG4, that do not substantially affect virulence in the mouse tail vein injection model. HIS1 and LEU2 have previously been demonstrated to be neutral for virulence (2, 22), and we show in this paper that ARG4 is also neutral. We also describe a new set of complementing plasmids with auxotrophic markers from heterologous Candida species, a fusion PCR technique for generating deletion constructs, and a protocol for sequential selection of deletions of the first and second alleles. Taken together, these materials and procedures greatly facilitate the generation of homozygous knockout mutants that are suitable for virulence studies.
The efficiency of these approaches was demonstrated by creating de novo disruptions of two previously characterized genes, TUP1 and EFG1, as well as a novel gene, RBD1. The well-described hyperfilamentation phenotype of the tup1Δ/tup1Δ mutant (9) and the well-documented hypofilamentation defect of the efg1Δ/efg1Δ strain (42) were recapitulated in the new strain backgrounds. Disruption of the previously uncharacterized open reading frame RBD1 resulted in a novel colony morphology phenotype; this phenotype was partially suppressed by introduction of a wild-type version of the disrupted gene.
Using the new method to disrupt TUP1 and EFG1, we observed a high frequency of targeted integration to the correct genomic locus, approximately 90% for the first allele and 70% for the second allele. We are not aware of the average efficiency of prevailing gene disruption methods, which are typically carried out with long PCR primers with 60 to 80 nucleotides of homology. In practice, however, it is common to observe a relatively low ratio (e.g., 1 in 100) of correct disruption products relative to other types of genetic changes, particularly in disruption of the second allele. There are several possible explanations for the improved efficiency of the new methods described here. First, 350 nucleotides of homology to regions flanking the target locus (as opposed to 60 to 80 nucleotides) likely facilitates homing to the correct locus, as has been reported for Cryptococcus neoformans (30). Second, since the two C. albicans alleles of a given gene are often heterologous in sequence (54), the opportunity for recombination at an exact sequence match between the disruption fragment and genomic DNA should be enhanced by inclusion of both allelic sequences in the disruption construct that is synthesized with genomic DNA as a template. Third, sequential gene disruption with different auxotrophic markers for each allele maintains selection for the first knockout at the time that the second is created, thereby avoiding selection of reintegrants into the first disrupted allele. Finally, the high efficiency of creating knockouts with selectable genes from related species instead of from C. albicans itself suggests that unwanted recombination events associated with standard C. albicans auxotrophic markers commonly involve recombination with the marker locus rather than the gene of interest.
Others have reported methods for creating large numbers of targeted gene mutations in C. albicans. For example, Roemer et al. presented the PCR-based GRACE (gene replacement and conditional expression) method, in which one allele of a target gene is replaced by standard methods and the promoter of the second allele is replaced with a regulatable promoter, with a dominant selectable marker (sat1) used to select the second event (36). This method has the advantage that expression of essential as well as nonessential genes may be disrupted; however, the efficiency of this method has not been reported. Likewise, Enloe et al. (14) described a method for creating homozygous gene disruptions in C. albicans with a single round of transformation followed by selection for homozygosity. This method relies on gene conversion or mitotic recombination (crossover) events to replace the second allele of the target gene with the disrupted copy and has the advantage of requiring only one round of transformation to disrupt both copies of the target gene. A disadvantage is that the gene conversion or mitotic recombination event that results in homozygosis of the deletion mutation will also affect adjacent chromosomal DNA, potentially including the entire chromosome arm distal to the recombination site. Thus, mutants created with this strategy may differ from the parental strain not only by the targeted deletion mutation but also by loss of heterozygosity along the affected chromosome. While this procedure is extremely useful for rapidly determining whether a given gene is essential, it cannot be used routinely to create genetically matched strains.
The new strains and procedures that we describe also provide an advantage in virulence studies by avoiding the use of URA3 as a selectable marker. As previously reported, ura3Δ/ura3Δ strains of C. albicans have striking virulence defects (Fig. (Fig.4A)4A) (22). Because URA3 expression can be subject to chromosome position effects, its use in genetic manipulations may produce unwanted effects on virulence. In contrast, as we show here, LEU2, HIS1, and ARG4 individually do not affect virulence in the tail vein injection model. Only the triple mutant, SN152, displayed significantly attenuated virulence, and this defect is relatively mild compared to that conferred by deletion of URA3 (Fig. (Fig.4).4). We conclude from these results that LEU2, HIS1, and ARG4 may be used safely to mark disrupted genes intended for virulence analysis since even if expression of marker genes is incomplete at heterologous loci, the aggregate effect on virulence should be no greater than that caused by the absence of the markers altogether. The mild virulence defect of the SN152 triple mutant is in keeping with the observation that GCN4, a general regulator of amino acid levels, is not required for virulence in C. albicans (8); i.e., the upregulation of amino acid biosynthetic pathways does not seem to be necessary for normal virulence. It unclear why URA3 is so critical for virulence, but a likely possibility is that nutritional components are differentially available in the context of the mouse host, with uridine being in particularly short supply.
Our recommendations for use of the new strains are as follows. SN152, with its three auxotrophic markers (his1, leu2, and arg4), offers a convenient background for most studies. The presence of a third auxotrophic marker is particularly useful for complementation analysis, in which a mutant phenotype can be reversed (at least partially) by addition of a wild-type copy of the deleted gene. Because SN152 has a mild virulence defect, however, it should probably not be used for assessing subtle virulence defects. For such uses, we recommend instead the fully virulent strains SN87 (his1 leu2) and SN95 (his1 arg4). In these strain backgrounds, linkage of a mutant phenotype can be correlated to deletion of the target gene either by examination of multiple independent knockout mutants or by adding back the wild-type version of the gene with a dominant selectable marker such as sat1 (36, 35). For studies focused on mutants with substantial virulence defects, SN152 may still be used, since the potential for poor expression of C. dubliniensis HIS1, C. dubliniensis ARG4, and C. maltosa LEU2 to attenuate virulence should be no greater than the mild defect of SN152 itself.
An advantage of an efficient gene disruption system is that, by reducing the time and effort required to make individual knockout mutants, large numbers of previously uncharacterized genes can be targeted. Screens of such mutants can yield rapid and unexpected insights. As an example, we disrupted RBD1, a gene that had not previously been studied in C. albicans. Surprisingly, this mutant yielded a novel “loss-of-center” colony morphology phenotype on Spider medium, almost opposite that of the well-studied tup1 mutant. Although we had no preconceived ideas that RBD1 might affect filamentation and, perhaps, cell-cell communication in C. albicans, these are now both reasonable areas for future research.
The past 10 years have produced exponential increases in publicly available sequence information for organisms known to cause disease in humans. Genome-scale studies and bioinformatics have already produced key scientific insights into mechanisms of pathogenesis (7, 13). The development of essential experimental tools such as standardized strains and efficient, rapid disruption technologies is crucial to test ideas developed from such studies. We have already used the new strains, reagents, and techniques presented in this article to construct 144 homozygous deletion mutants of C. albicans within a few months' time (S. M. Noble and A. D. Johnson, unpublished data). Performed on a larger scale, it should be possible to create a C. albicans genomic disruption library (of nonessential genes) within a year or two. Such a genome-scale knockout collection for C. albicans would enable systematic, direct screens for virulence determinants of this important fungal pathogen.
We are grateful to B. B. Magee and P. T. Magee for the gift of RM1000#2, to P. T. Magee and E. P. Rustchenko for communication of unpublished results, to A. M. Ramon and W. A. Fonzi for plasmid pLUBP, to G. R. Fink for plasmid pBES116, to O. Liu and H. D. Madhani for protocols and helpful discussions regarding fusion PCR, and to the Stanford Genome Technology Center for provision of the C. albicans genome sequence. We are also grateful to H. D. Madhani for careful reading of the manuscript.
S.M.N. was a Howard Hughes Postdoctoral Research Physician during the period of this research. The work was supported by a grant (RO1 AI49187) from the National Institutes of Health.