Candida albicans gains azole resistance by altering sphingolipid composition

Fungal infections by drug-resistant Candida albicans pose a global public health threat. However, the pathogen’s diploid genome greatly hinders genome-wide investigations of resistance mechanisms. Here, we develop an efficient piggyBac transposon-mediated mutagenesis system using stable haploid C. albicans to conduct genome-wide genetic screens. We find that null mutants in either gene FEN1 or FEN12 (encoding enzymes for the synthesis of very-long-chain fatty acids as precursors of sphingolipids) exhibit resistance to fluconazole, a first-line antifungal drug. Mass-spectrometry analyses demonstrate changes in cellular sphingolipid composition in both mutants, including substantially increased levels of several mannosylinositolphosphoceramides with shorter fatty-acid chains. Treatment with fluconazole induces similar changes in wild-type cells, suggesting a natural response mechanism. Furthermore, the resistance relies on a robust upregulation of sphingolipid biosynthesis genes. Our results shed light into the mechanisms underlying azole resistance, and the new transposon-mediated mutagenesis system should facilitate future genome-wide studies of C. albicans.

C andida albicans is a major fungal pathogen of humans, affecting millions of people and kills hundreds of thousands worldwide annually 1 . Mortality rates of invasive C. albicans infection remain high despite the treatment with existing antifungal therapies. Multidrug-resistant Candida species and strains are rapidly emerging and spreading globally, threatening to render our entire antifungal arsenal obsolete 2 . Azoles are common first-line drugs against most Candida species in systematic infections, which act by inhibiting the 14α-demethylase Erg11 in the ergosterol biosynthesis pathway and causing the accumulation of the toxic sterol 14,24-dimethylcholesta-8,24(28)dien-3β,6α-diol (DMCDD) that permeabilizes the plasma membrane 3 . Extensive research has demonstrated that fungal pathogens can quickly evolve various resistance strategies after exposure to an antifungal agent. C. albicans can gain resistance to azoles via genetic alteration of the drug target Erg11, upregulation of efflux pumps Cdr1, Cdr2, and Mdr1, and inactivation of Erg3 that synthesizes the toxic sterol 4,5 . However, our mechanistic understanding of antifungal resistance is incomplete 2 . It remains unclear why some fungal species are intrinsically resistant or easily acquire resistance to multiple antifungal drugs. Thus, it is imperative to identify and understand the molecular mechanisms that govern antifungal resistance.
Genome-wide genetic screening is a powerful tool to identify genes involved in a biological process. However, it has been difficult to apply these technologies to C. albicans due to its diploid genome. Although several recent studies conducting large-scale genetic screens of C. albicans mutant libraries resulted in important findings 6-10 , investigation of many aspects of C. albicans biology still relies on inference from data obtained in the studies of Saccharomyces cerevisia 11,12 . This approach inevitably misses out genes that perform specific functions and determine unique traits in C. albicans. For example, S. cerevisiae is nonpathogenic and does not grow true hyphae which are the predominant morphology of C. albicans in host tissues and a key virulence factor 11 . Currently, most C. albicans genes with limited homology to S. cerevisiae remains uncharacterized 13 .
Recently, some haploid C. albicans strains were isolated 14 . Although avirulent, the haploids possess most traits that define the species including the ability to undergo the yeast-hyphae transition, the white-opaque phenotypic switch, the formation of chlamydospores, and mating 14 . The haploids have opened up opportunities for scientists to explore large-scale, genome-wide genetic screening strategies in this pathogen. The recent application of CRISPR-cas9 technology in haploid C. albicans in the identification of new virulence regulators and genetic networks demonstrates the massive potential of using haploid C. albicans in genetic screens 15 .
Transposon-mediated mutagenesis is a widely used genetic tool. Transposons are movable genetic units and have two essential components, a transposase and a pair of transposonspecific inverted terminal repeats (ITRs). One example is the piggyBac (PB) transposable element, first isolated from the moth Trichoplusia ni and later adapted to other organisms from bacteria to mammals 16,17 . PB transposase recognizes the ITRs and moves the entire content from one chromosomal locus and integrates it at a TTAA motif in a new locus 18 . PB transposition occurs via a precise cut-and-paste mechanism which leaves no footprints behind, resulting in a full restoration of the donor site 19 . Thus, a phenotype caused by a PB insertion can be easily verified by phenotype reversion. Furthermore, PB transposon inserts randomly throughout the genome 20 .
In this study, we develop a highly efficient, PB-based mutagenesis system using stable haploid C. albicans strains. We apply this technology to conduct genome-wide genetic screens for genes involved in resistance to the antifungal drug fluconazole and discover a new resistance mechanism.

Results
Construction of the piggyBac transposition system in C. albicans. We first optimized the piggyBac transposase (PBase) gene for use in C. albicans. We replaced all CUG codons with UUG because CUG encodes serine instead of the conventional leucine in C. albicans, added a 3×SV40 nuclear localization signal to its N-terminus, and controlled the expression using the tetracyclineinducible (Tet-On) promoter (Fig. 1a). Also, we mutated seven amino-acid residues to generate a hyperactive transposase according to previous studies 21 . We then integrated this CaPBase gene, using SAT1 as the selectable marker, at the ADH1 locus of the stable, uridine auxotrophic (Ura-) haploid C. albicans strain GZY803 22 to generate the YW01 CaPBase-expressing strain.
To test whether PB transposition would occur in C. albicans, we transformed the donor plasmid pPB[URA3] carrying the ITR-URA3-ITR cassette into YW01 cells that had been grown in the absence or presence of doxycycline (Dox) for 48 h and then spread the cells onto uridine-deficient (-Uri) plates. The transformation of Dox-induced cells led to the growth of numerous colonies, while the uninduced cells produced only a few likely due to the basal expression of CaPBase (Fig. 1b), an event that happened in <0.001% of the cells. This result indicated that the Dox-induced CaPBase excised the PB cassette from the plasmid and integrated it into the genome.
Previous studies in other organisms showed that PB excision did not leave any footprint behind 19 . To confirm this characteristic in C. albicans, we constructed a C. albicans strain in which we inserted the PB[URA3] cassette at a TTAA site (n.t. 709-712) within the open reading frame (ORF) of ARG4 (Fig. 1c), yielding YW02 that is arginine auxotrophic (Arg−) and uridine prototrophic (Ura+) and also expresses CaPBase from the Tet-On promoter at the ADH1 locus. Precise excision of PB would convert YW02 cells from Arg− to Arg+. After Dox induction for 48 h, we spread the YW02 cells onto glucose minimal medium (GMM) plates, picked 20 independent colonies, and confirmed that all 20 clones grew well on -Arg-Uri plates (Fig. 1d), indicating that the cells had excised PB[URA3] out of ARG4 precisely. This was confirmed by site-specific PCR using two pairs of primers as shown in Fig. 1c (Fig. 1e shows examples of 5 clones) and DNA sequencing analysis of the PCR products. To map PB insertion sites, we performed inverse PCR as described in Fig. 1f which allowed amplification of the DNA fragment encompassing a transposon junction. DNA sequencing analysis of the PCR products and homology search of the Candida Genome Database (CGD) located the PB insertion sites on all eight chromosomes (Supplementary Table 1), seven within ORFs and 13 in intergenic regions (Fig. 1g). Our results demonstrated random PB transposon insertion and precise excision in C. albicans.
Ploidy and PB transposition efficiency analysis. Autodiploidization observed in C. albicans haploids has been a major concern regarding their applicability in genetic screens 14 . To evaluate ploidy stability during transposition, we first performed qPCR analysis to measure CaPBase expression levels at timed intervals in YW02 cells grown on solid media with or without Dox. The results demonstrated a sharp increase of CaPBase expression upon Dox addition (Fig. 2a). In parallel, we examined cell ploidy by flow cytometry and found that the cells retained 1N DNA content for at least 72 h (Fig. 2b) regardless of Dox addition, indicating a stable ploidy during the robust expression of CaPBase. Tests in liquid medium produced similar results (Supplementary Figure 1a and 1b).
We next estimated PB transposition frequency. Aliquots of YW02 cells grown on solid medium containing Dox were harvested at intervals, and each was divided into two equal parts, and one was spread onto a YPD plate that allows all cells to grow and the other onto GMM+histidine (His) plates where only Arg +Ura+ transposants can grow. Transposition frequency was calculated by dividing the number of colonies on the GMM plate with that on the YPD plate. Figure 2c shows that PB transposition frequency increased with time during Dox induction, reaching~5 % at 72 h. We also achieved a similar frequency in liquid medium (Supplementary Figure 1c).
Multiple transposon insertions in a genome complicate subsequent identification of the mutation responsible for a phenotype. Although a single copy of PB transposon exists at the ARG4 locus in YW02, the copy number may increase when transposition happens in S or G2 cells 23

ARTICLE
qPCR to determine the level of PB DNA in comparison with a control gene CDC28 in the genome. All clones were found to contain a single copy of PB ( Fig. 2d and Supplementary  Figure 1d), indicating that the vast majority of transposants harbored a single PB insertion in the genome.
Genome-wide analysis of PB transposon distribution. Next, we analyzed PB distribution in the genome using next-generation sequencing (NGS  and subjected to NGS. We acquired 11,595,368 overlapping read pairs, each containing one part matching the PB element and the other unambiguously matching a sequence of the C. albicans genome. The PB transposon distribution in 10-kb non-overlapping windows is largely uniform throughout the genome ( Fig. 2e and Supplementary Figure 1e). On average, there were 20 unique insertion sites in a 10-kb window (Fig. 2f), i.e.,~1 every 500 bp. We detected PB inserts in nearly 5000 genes out of a total of 6466 annotated and predicted genes in CGD. The failure to find PB insertion in the rest of the genes is likely due to their essentiality or the fitness cost caused by the PB insertion, consistent with the~1100 essential genes in S. cerevisiae 24 . Nucleotide motif analysis of the insertion sites revealed 82.4% at TTAA sites, 4.5% at TTAG, 4.4% at CTAA, 3.6% at TTGT, and 5.1% at other non-specific sites (Fig. 2g). Also, we found significant enrichment of Ts and As in the 5 nucleotides immediately upstream and downstream of the PB inserts, respectively (Fig. 2g). PB insertion occurred more frequently in promoter regions than in ORFs. This could be due to the enrichment of TA dinucleotides in promoters, which facilitates the unwinding of double-stranded DNA during transcription initiation. Another reason could be the fitness cost caused PB insertion in some ORFs, which may reduce their presence in the transposant population.
PB mutagenesis for forward genetic screens of 5-FOA resistant mutants. To evaluate the effectiveness of the PB transposon system in genetic screens in C. albicans, we performed a proof-ofprinciple screen for mutants resistant to 5-fluoroorotic acid (5-FOA). As 5-FOA is converted to the toxic 5-fluorouracil in strains expressing a functional URA3 25 , we wanted to determine how frequently we can identify ura3 mutants in the transposon insertion library. We constructed a starting strain YW05 by using URA3 to delete HIS1 in GYZ803 and then inserting a PB[HIS1] cassette in ARG4 ORF. YW05 also expresses CaPBase from the ADH1 locus. Figure 3a describes the steps of the screen. YW05 cells were grown as patches on YPD plates containing Dox for 48 h to induce the transposition. Then, the cells were replicatransferred to GMM+Uri plates to enrich Arg+His+ transposants which were then transferred to GMM+Uri+5-FOA plates to select for 5-FOA-resistant mutants. Among 30 colonies randomly picked from the 5-FOA plates were 26 true 5-FOAresistant mutants (Fig. 3b) and 4 clones (#3, #7, #18, and #21) were false and thus excluded from further characterization. Colony PCR using a pair of primers flanking URA3 failed to detect the 1.4-kb amplicon expected from the WT strain in 22 of the 26 5-FOA-resistant mutants (Fig. 3c), suggesting PB[HIS1] insertion in URA3. URA3 seemed intact in four clones, #4, #9, #15, and #26, suggesting PB disruption of other genes leading to 5-FOA resistance (Fig. 3c). Next, we performed PCR analysis of the 22 ura3 mutants by pairing a URA3-flanking primer, URA3F or URA3R, with the PB-specific PBLR primer (Fig. 3d), followed by DNA sequencing analysis of the PCR products. We identified 4 different insertion sites in the URA3 locus (Fig. 3e). One was located in the promoter region, and all the others were located in the ORF. Also, 19 PB insertions occurred at TTAA, one at TTAG, and two at CTAA motifs.
We also performed an NGS comparison of the mutant pools before and after 5-FOA treatment. We obtained 3,266,688 and 3,865,123 read pairs of PB insertion junctions from 5-FOAtreated and untreated samples respectively. Ranking the fold changes of genes after 5-FOA treatment, URA5 and URA3 occupied the first and second positions (Fig. 3f), exceeding the third by~10 times. URA5 and URA3 also occupied the top eight places in the rank of fold changes of insertion sites after 5-FOA treatment (Supplementary Figure 2a). Consistently, inverse PCR and sequencing analysis of 5-FOA resistant clones #4, #9, #15, and #26 confirmed PB[HIS1] insertion at the same position in URA5 ORF (Fig. 3g). Deletion of URA5 was found previously to cause partial resistance to 5-FOA in S. cerevisiae 25 . However, deleting URA5 in both haploid and diploid C. albicans caused comparable 5-FOA resistance as deleting URA3 (Supplementary Figure 2b). Ura3 and Ura5 catalyze two consecutive steps in the de novo biosynthesis of pyrimidines 26 . Taken together, this proof-of-principle screen demonstrated the superb power of this PB-mediated mutagenesis system in identifying genes required for a cellular process.
Genome-wide genetic screening for fluconazole-resistant genes.
To identify new antifungal resistance mechanisms, we applied the PB-transposon mutagenesis system to perform genome-wide profiling of genes whose inactivation causes resistance to fluconazole. The rationale of the screen is: if PB insertion into a gene causes resistance to fluconazole, cells carrying this mutation will outcompete sensitive cells when grown in the presence of fluconazole and become overrepresented in the population. Thus, the subsequent NGS reads of PB insertions in this gene in the pooled genomic DNA will increase correspondingly. The PB insertion library was prepared by first growing YW02 cells in YPD+Dox medium and then dividing the culture into two equal parts for further cultivation in Dox-free medium with or without fluconazole. Cells were harvested after doubling 10 times and processed for NGS analysis.
We detected, in three independent experiments, an average of 5250 and 5122 genes from fluconazole-treated and untreated samples and ranked the genes according to the levels of enrichment after fluconazole treatment. Among the top 10 were three genes that either had been linked to azole susceptibility or are involved in ergosterol biosynthesis including ERG3, ERG6, and ERG251 ( Fig. 4a and Supplementary Data 2). We randomly selected 10 genes ranked higher than PMC1, which is the lowest ranking known resistant genes in our list, to construct haploid deletion mutants and found that they all showed increased resistance to fluconazole (Fig. 4b). Gene Ontology (GO) analyses of the genes that showed a significant increase in abundance in fluconazole-treated cells (log2 fold change > 1 and p-value < 0.05) revealed an overrepresentation of genes associated with ergosterol and sphingolipid biosynthesis (Fig. 4c). We also performed Gene Set Enrichment Analysis (GSEA) and found significant enrichment of genes associated with ergosterol and sphingolipid biosynthesis in fluconazole-treated cells in the leading edge subset that has the highest correlation with fluconazole resistance (Fig. 4d). The latter group included FEN1, FEN12, and ARV1 that play important roles in sphingolipid biosynthesis and trafficking 27,28 . Like ergosterol, sphingolipids are the main components of the fungal plasma membrane which is both the target of azole and polyene classes of antifungal drugs and site of drug transporters. We reasoned that the discovery of these genes in our screen likely reflects some previously unknown mechanisms of drug resistance. Thus, we further characterized FEN1 and FEN12, both encoding fatty-acid elongases that synthesize verylong-chain fatty acids as precursors of sphingolipid biosynthesis 28 . We verified that fen1Δ/Δ and fen12Δ/Δ mutants constructed in the diploid BWP17 background also exhibited significantly higher resistance to fluconazole than the WT strain (Fig. 4e).
fenΔ/Δ mutants can tolerate higher levels of the toxic sterol. Since blocking DMCDD synthesis by inactivating ERG3 is a known resistance mechanism to fluconazole 29 , we quantified ergosterol and DMCDD in whole cell lysates and the plasma membrane using LC-MS. Due to the lack of standard DMCDD, we included an erg3Δ strain to ensure correct identification of this molecule by combining the detected molecular mass and the absence of the compound in erg3Δ cells (Fig. 5a, top and Supplementary Figure 3). LC-MS detected a~90% decrease of ergosterol in the whole cell lysates of both fen1Δ/Δ and fen12Δ/Δ mutants compared to WT controls grown in drug-free media (Fig. 5a, middle), indicating that lacking either FEN1 or FEN12 impairs ergosterol biosynthesis severely. As expected, fluconazole treatment caused a dramatic drop of ergosterol levels in all strains. Unexpectedly, we detected a~270% increase in the amount of DMCDD in both fen1Δ/Δ and fen12Δ/Δ cell lysates compared with WT cells upon fluconazole treatment (Fig. 5a, bottom). However, the DMCDD level in the plasma membrane fraction of fen1Δ/Δ and fen12Δ/Δ cells was found to be 35 and 29% of that in WT cells (Fig. 5a, bottom). The results suggest that, unlike erg3Δ/Δ, the higher fluconazole resistance of fenΔ/Δ mutants is not due to decreased production of the toxic sterol.
Significant changes in sphingolipid composition in fenΔ/Δ mutants. As ergosterol and DMCDD insert themselves into the cell membrane mainly through interaction with sphingolipids 30 , we speculated that changes of the sphingolipid profile in fen1Δ/Δ and fen12Δ/Δ might be responsible for the tolerance of DMCDD. To test this hypothesis, we performed lipidomics profiling by LC-MS. We detected rather different fatty-acid profiles between fen1Δ/Δ and fen12Δ/Δ mutants ( Fig. 5b and Supplementary  Figure 4). The levels of fatty acids with 24 and 26 carbons decreased by 74-85% in both mutants grown in drug-free media (Fig. 5b), consistent with the loss of the very-long-chain fatty acid elongase. Varying degrees of increase or decrease were observed in the levels of fatty acids with 14-22 carbons; for instance, fen1Δ/ Δ cells grown in drug-free media showed 10-60% decrease in a b Induction YPD + Dox GMM + Uri GMM Clones 1 2 5 6 10 11 12 13 14 16 19 22 23 24 25 27 28 8 17 20 29 (Fig. 5b). The results suggest that increasing the cellular levels of some species of sphingolipids might be a natural stress response to fluconazole. We speculate that high levels of sphingolipids could improve cell membrane strength leading to higher tolerance of the toxic sterol.   Fig. 4 Genetic screens for fluconazole-resistant mutants. a Volcano plot of fold changes of genes with PB inserts in PB transposition mutants grown in media containing 100 µg/mL fluconazole in comparison with the same mutants grown in normal media. p-Values were calculated using the Wald test. Ten candidates selected for phenotype confirmation are indicated. Genes known to be associated with fluconazole resistance and ergosterol and sphingolipid biosynthesis are shown with different colors. b Gene deletion of 10 candidates randomly selected from the top-ranked genes in a confirms increased resistance to fluconazole. Fluconazole susceptibility was evaluated using the standard CLSI broth microdilution protocol 67 . Cultures were incubated for 24 h, and growth was measured as OD 600 and expressed as a percentage of the growth in the control wells with no drug (DMSO alone) for each strain. c Gene Ontology (GO) analysis of the PB-mutated genes that led to statistically significant fold increases (p < 0.05) after fluconazole treatment. Mechanisms of sphingolipid overproduction in fenΔ/Δ mutants. We further investigated the mechanisms behind the fluconazole-induced sphingolipid overproduction. Fungal cells possess a remarkable capacity to counterbalance a disturbance to maintain cellular homeostasis. In response to fluconazole treatment, C. albicans reacts by overexpressing UPC2 which encodes a transcription factor that regulates ergosterol biosynthesis (Fig. 6a summarizes the biosynthetic pathways of ergosterol and sphingolipids) and sterol uptake 31 . Indeed, we detected the overexpression of UPC2, ERG3, and ERG11 in both WT and fenΔ/Δ strains after fluconazole treatment (Fig. 6b). UPC2 also regulates genes involved in sphingolipid biosynthesis [32][33][34][35] . Consistently, we detected overexpression of the ceramide synthase gene LAG1, the IPC synthase gene AUR1, the MIPC synthase gene MIT1, and the M(IP) 2 C synthase gene IPT1 upon fluconazole treatment (Fig. 6b). Importantly, the fluconazole-induced UPC2 overexpression is much stronger in fenΔ/Δ cells than in WT cells. The results are consistent with the higher amounts of sphingolipids found in the mutants grown in the presence of fluconazole.
To determine whether the upregulation of sphingolipid biosynthesis is responsible for the increased resistance to fluconazole in fen1Δ/Δ and fen12Δ/Δ mutants, we attempted to delete UPC2 and LAG1 in fen1Δ/Δ and fen12Δ/Δ strains. Despite repeated efforts, we could not delete both copies of LAG1 in either mutant and those of UPC2 in fen1Δ/Δ, suggesting synthetic lethality. However, we found that deleting one copy of LAG1 or one copy of UPC2 in fen1Δ/Δ and fen12Δ/Δ appeared to increase their sensitivity to fluconazole (Fig. 6c). Also, deleting both copies of UPC2 in fen12Δ/Δ abolished its resistance to fluconazole (Fig. 6c). The results suggest that Upc2-induced upregulation of sphingolipid biosynthesis may underlie the increased fluconazole resistance of fen1Δ/Δ and fen12Δ/Δ mutants. Consistently, the fluconazole-induced upregulation of ERG3, ERG11, LAG1, AUR1, MIT1, and IPT1 was significantly reduced or abolished when one or both copies of UPC2 was deleted (Fig. 6b). Together, the results corroborate the hypothesis that fen1Δ/Δ and fen12Δ/Δ mutants increase the cellular levels of sphingolipids in response to fluconazole by strongly upregulating the expression of genes a b   Error bars represent s.d. from the mean of three independent experiments. Significance was measured using unpaired t-test (*p < 0.05, **p < 0.01, ***p < 0.001, n.s., not significant). c Fluconazole susceptibility of WT and the indicated mutants was evaluated using the standard CLSI broth microdilution protocol. Following 24 h of incubation, growth was measured as OD 600 and expressed as a percentage of the growth in the control wells with no drug (DMSO alone) for each strain involved in sphingolipid biosynthesis, leading to tolerance of the toxic sterol.

Discussion
In recent years, considerable efforts have been made to introduce transposon mutagenesis strategies to create libraries of heterozygous 36,37 or homozygous 38,39 mutants in C. albicans. In general, these methods start with in vitro transposition of a transposon-containing cassette, Tn5-UAU1 or Tn7-UAU1, into a C. albicans genomic DNA library 36,37 . Then, the mutagenized library is transformed into Escherichia coli for amplification before transformation into C. albicans. These methods are tedious, labor-intensive, and unsuitable for high-throughput genome-wide genetic studies. Also, their use in diploid C. albicans creates heterozygous mutations that do not result in readily detectable phenotypes except for a small number of haploinsufficient genes 40,41 . Mielich et al. recently introduced the Activator/Dissociation transposon of maize into haploid C. albicans 42 , but the system is not very useful at its current state because of frequent autodiploidization.
Here, we introduce a simple and efficient PB transposon-based mutagenesis system in haploid C. albicans, which has the following advantageous features. First, a single transposon insertion in the promoter or ORF of a gene is, in most cases, sufficient to inactivate a gene and produce a phenotype. Second, the transposon donor is inserted into a selectable marker gene, i.e., ARG4, and the transposase is expressed from a different locus under the control of the Tet-On promoter. Thus, a mutant library can be generated merely by growing the cells in a medium containing Dox to express the transposase and then selecting Arg+ transposants in -Arg media. Third, transposition occurs at high frequencies and in a random manner. Our NGS analysis showed an even distribution of PB insertions on all chromosomes and identified PB inserts in~5000 genes (Fig. 2e). Fourth, each mutant cell contains a single PB insertion, which allows straightforward mapping of the affected gene by inverse PCR and DNA sequencing. Fifth, the precise excision of PB from the insertion site restores the coding sequence allowing easy confirmation of the phenotype-genotype relationship. Our successful identification of URA3 and URA5 in 26 out of 30 random 5-FOAresistant mutants demonstrates the superb power of the PB transposon mutagenesis system (Fig. 3). Furthermore, the PBmediated mutagenesis system is readily applicable to wholegenome epistasis analysis of gene networks by first deleting a gene of interest before activating the transposition and screening for synthetic phenotypes. We anticipate that this PB transposonbased mutagenesis platform will significantly accelerate discoveries to advance our understanding of the biology and pathogenesis of this pathogen.
Frequent isolation of multidrug-resistant Candida worldwide is a cause for concern because currently only three classes of drugs are available approved for the treatment of systemic fungal infections 43,44 . Hard lessons learned since the discovery of penicillin have taught us that microbial pathogens are extremely innovative in quickly finding tricks to defeat every one of our therapies. To win the war against infectious disease, it is imperative to elucidate how a pathogen has acquired resistance and even to predict how it can do it. The development of the PBtransposon mutagenesis system has allowed us to conduct genome-wide screening in C. albicans to identify genes underlying antifungal susceptibility. Our NGS analyses of PB transposon insertion sites in mutants selected by growth in the presence of fluconazole in comparison with mutants grown in drug-free media revealed significant enrichment of genes involved in ergosterol biosynthesis, in particular, ERG3 and ERG6 whose mutation is well known to increase fungal resistance to azoles [45][46][47][48] . Deleting any one of 10 genes randomly picked from our resistant mutant list all led to increased fluconazole resistance (Fig. 4b), indicating that the screen was successful. Among the highest ranking genes were FEN1, FEN12, and ARV1 which play important roles in sphingolipid biosynthesis 49,50 . The understanding that sphingolipids are major components of fungal cell membrane and interact both physically and functionally with ergosterol 30,51,52 , the target of azole and polyene antifungal drugs, prompted us to conduct a comprehensive analysis of FEN1 and FEN12. Our first informative finding was a severalfold increase in the cellular levels of the toxic sterol in both fenΔ/Δ mutants compared to WT cells (Fig. 5a), which indicated a resistance mechanism distinct from that of the erg3 mutant 2,46 . Interestingly, the level of the toxic sterol in the plasma membrane is significantly lower in the mutants than in the wild type (Fig. 5a), suggesting that the mutants pack fewer toxic sterol molecules into the membrane. Furthermore, we detected dramatic changes in the sphingolipid composition in whole cell extracts (Fig. 5b). The most striking change is the increase of more than 3 orders of magnitude of some MIPC species in fen12Δ/Δ cells upon fluconazole treatment, which also increased the total cellular level of sphingolipids. An illuminating observation was the sharp increase of the same MIPC species in WT cells after fluconazole treatment (Fig. 5b), indicating that C. albicans alters its sphingolipid composition as a natural stress response mechanism to the disruption of ergosterol biosynthesis. This could be a conserved survival strategy because eukaryotic cells are known to have a huge capacity to preferentially modify the cellular levels and make-up of sphingolipids in response to changes in sterol composition in their membrane as an adaptation mechanism to maintain membrane structure and function 52 . In S. cerevisiae, mutation of ELO3, a homolog of FEN12, also affected sphingolipid composition significantly 53 . Many previous studies have revealed intricate relationships between ergosterol and sphingolipid biosynthesis and resistance to azoles. Particularly consistent with our findings, Prasad et al. found that blocking M(IP) 2 C synthesis by deleting the IPT1 gene significantly increased C. albicans susceptibility to azoles 54 ; furthermore, the level of the drug efflux pump Cdr1 in the plasma membrane was reduced in ipt1Δ/Δ cells as well as in cells treated with fumonisin B1, an inhibitor which blocks the synthesis of phytoceramide 55 . Consistently, we demonstrated that blocking sphingolipid biosynthesis reduced the fluconazole resistance in fen1Δ/Δ and fen12Δ/Δ cells. Sphingolipids are abundant components of the fungal cell membrane and play both structural and signaling roles 56 . Numerous experiments have shown that changes of sphingolipid composition can dramatically alter the properties of cell membranes, such as rigidity, fluidity, and domain organization which further influences cellular functions 57 . Thus, the marked changes of the cellular sphingolipid composition in fenΔ/Δ mutants might lead to fluconazole resistance by reducing the incorporation of the toxic sterol into the cell membrane and limiting subsequent structural and functional damage of the cell membrane. Although we do not have direct evidence confirming repellence of the toxic sterol from the cell membrane by high levels of sphingolipids, this can be expected based on the chemical properties of sterols and sphingolipids and is indirectly supported by various previous studies. Cell membranes have a complex organization featuring domains with distinct lipid compositions known as rafts whose formation is determined by the hydrophobic interaction between sterols and sphingolipids 58,59 . However, the toxic sterol is structurally different from ergosterol in terms of the absence of B-ring unsaturation, the presence of the 6-OH group and the additional C-14 methyl group, and side-chain modification. In particular, the polar 6-OH group is thought to interfere with sterol-sphingolipid packing in the plasma membrane, causing membrane damage 60,61 . The high level of sphingolipids in fenΔ/Δ mutants could strengthen the cell membrane by restraining membrane deformation. When the additional free energy required for incorporating polar groups of the toxic sterol cannot be compensated by membrane deformation, they must be pushed out to reach a lower energy status. Thus, more toxic sterol molecules with a polar side chain would be repelled from the membrane by the nonpolar chain of sphingolipids. Taken together, it is reasonable to believe that high levels of sphingolipids can repel toxic sterol molecules from the cell membrane. It was reported that fen1Δ/Δ and fen12Δ/Δ mutants exhibited increased sensitivity to amphotericin B 62 . It is widely accepted that amphotericin B binds to ergosterol in a parallel manner to form barrel-stave type pores and penetrate cell membranes, in which their hydrophilic polyhydroxy side is pointing inward to constitute the pore lining and their hydrophobic lipophilic heptaene part is directing outward to interact with the membrane interior 63,64 . While the high levels of sphingolipids in the membrane hinder the insertion of the hydrophilic toxic sterol, the binding of amphotericin B to ergosterol is not affected; and on the contrary, the shorter sphingolipids in fenΔ/Δ mutants could expose ergosterol molecules and thus facilitate the binding to amphotericin B, making mutant cells more sensitive to this drug. In future studies, it would be interesting to know whether high cellular levels of sphingolipids contribute to azole resistance in clinical isolates of C. albicans and other fungal pathogens.
Plasmid construction. pPB[URA3] was constructed as follows: the URA3 marker was PCR-amplified from a pBluescript-based plasmid with primers URA3ApaI and URA3SacII (all primers used in this study are shown in Supplementary Data 1 Inverse PCR for mapping PB insertion sites. Genomic DNA was prepared using MasterPure Yeast DNA Purification Kit (Epicentre Technologies, MPY80200) and digested with BfuCI overnight. After heat inactivation of BfuCI, the digested genomic DNA was self-ligated with T4 DNA ligase at 16°C overnight. Primers used to recover the flanking sequence of the left side of the PB transposon were PBLf and PBLr. PCR products were purified from agarose gel and sequenced using Seq1 as the sequencing primer. The sequencing results were used to BLAST-search Candida Genome Database (www.candidagenome.org) to determine the PB insertion sites.
Quantitative PCR. To determine the copy number of PB[URA3], genomic DNA was purified using MasterPure yeast DNA purification kit and analyzed on STRATAGENE Mx3000p using Maxima SYBR Green qPCR Master Mix (2X) with separate ROX vial (Thermo Fisher Scientific, K0253) according to the manufacturer's instructions. PCR was carried out with primers CDC28 sense and CDC28 antisense to amplify CDC28, URA3 sense and URA3 antisense to amplify PB [URA3]. PB copy number was determined by ΔΔCT method using the genomic DNA of strain YW02 as a control, which has a single copy of PB 66 . Since the ratios of PB versus CDC28 were usually not integral numbers, we estimated the copy number to be the integral closest to the ratios.
To detect the expression of CaPBase, ERG11, ERG3, UPC2, LAG1, AUR1, MIT1, and IPT1, total RNA was isolated using the RNeasy Mini Kit (QIAGEN, 74106) and RNase-Free DNase Set (QIAGEN, 79254) and then reverse-transcribed using the SuperScript™ II Reverse Transcriptase (Invitrogen, 18064014). Changes in transcript levels of target genes were analyzed using the Maxima SYBR Green qPCR Master Mix (2X) with separate ROX vial with primers PBase sense and PBase antisense for CaPBase, ERG11 sense and ERG11 antisense for ERG11, ERG3 sense and ERG3 antisense for ERG3, UPC2 sense and UPC2 antisense for UPC2, LAG1 sense and LAG1 antisense for LAG1, AUR1 sense and AUR1 antisense for AUR1, MIT1 sense and MIT1 antisense for MIT1, IPT1 sense and IPT1 antisense for IPT1, and normalized to CDC28 transcript levels using the ΔΔCT method.
Transposition induction. In the experiments using integrated PB[URA3] or PB [HIS1] as the transposon donor, a single colony of YW02 or YW05 was first cultured in liquid YPD medium at 30°C overnight. Then, the cells were transferred to liquid or solid YPD medium supplemented with 50 μg/mL Dox to induce the expression of CaPBase and undergo transposition. To calculate transposition efficiency, aliquots of cells were harvested at timed intervals and spread on GMM plates to count the number of cells that had undergone transposition and on YPD plates to obtain the total number of cells. Transposition efficiency was calculated as the number of colonies on GMM plates divided by the number of colonies on YPD plates.
Flow cytometry. Flow cytometry analysis was performed using a BD FACSCalibur as described previously 14,22 . Mid-log phase cells were collected, washed, and fixed with 70% (vol/vol) ethanol (Sigma, E7023-500ML). Cells were washed with solution I (200 mM Tris-HCl, pH 7.5 and 20 mM EDTA) once and treated with 10 mg/ mL RNase A (Sigma, R4875-100MG) (1:100 dilution in solution I). Cells were washed with PBS and resuspended in 5 mg/mL propidium iodide (Sigma, 81845-25MG) (1:50 dilution in PBS) incubated in the dark at 4°C overnight. Stained cells were collected and resuspended in PBS and sonicated before loading.
Mapping PB insertion sites by next-generation sequencing. A single colony of YW02 or YW05 cells were inoculated into 5 mL of YPD and grown at 30°C to the exponential phase. Eighteen equal aliquots of the culture, each containing about 10,000 cells, were spotted on a YPD plate containing 50 µg/mL Dox to induce CaPBase expression. After 24 h of incubation at 30°C, cells on the induction plates were replica-transferred to a GMM plate without Dox to enrich the cells that had undergone transposition. The GMM plate was incubated at 30°C for 2 days before collecting the cells for the extraction of genomic DNA using the MasterPure Yeast DNA Purification Kit. Tagmentation was carried out by using the TruePrep DNA Library Prep Kit V2 for Illumina (Vazyme Biotech, TD501) according to the manufacturer's instructions. The first round of PCR was carried out to amplify PB together with its flanking sequences using a PB-specific primer close to the insertion site, PBseq and a primer annealing to the Tn5 adaptor, N7XX. The cycling parameters were: 3 min at 72°C; 30 s at 98°C; 12 cycles of 15 s at 98°C, 30 s at 60°C, and 3 min at 72°C; 10 min at 72°C; 4°C forever. The second round of PCR was performed using NEBNext® Ultra™ II Q5® Master Mix (NEB, M0544L) to add sequences required for Illumina sequencing with an Illumina flowcell attachment primer, TruPBseq and partial Tn5 adaptor primer, Illumina-R. Cycling conditions are: 1 min at 98°C; 7 cycles of 10 s at 98°C, 75 s at 67°C; 5 min at 67°C; 4°C forever. The size selection and purification of the PCR products were carried out with AMPure XP beads (Beckman Coulter, A63882).
Genetic screens using PB-mediated mutagenesis system. In the screen for 5-FOA-resistant mutants, insertional mutagenesis in YW05 cells was induced by culturing on YPD plate containing Dox for 24 h and then replica-transferred to GMM plates containing uridine to select for Arg+His+ cells. After incubation at 30°C for 24 h, cells on the above GMM plates were replica-transferred to GMM plates supplemented with uridine and 5-FOA. The 5-FOA plates were then incubated at 30°C until resistant colonies had formed. For screening in the liquid medium, Arg+His+ mutant cells were collected and washed three times with PBS. Two equal aliquots of cells were cultivated further in 20 mL GMM medium containing uridine with or without 5-FOA at 30°C for 48 h; then cells were harvested and processed for NGS analysis.
In the screen for fluconazole-resistant mutants, transposition-mediated mutagenesis was performed using strain YW02 grown in YPD plate containing Dox for 24 h at 30°C. Then, the cells were replica-transferred to GMM plates and incubated at 30°C for 24 h to enrich Arg+Ura+ cells that had undergone the transposition. Then, mutant cells were collected and washed three times with PBS. Approximately 2-4 × 10 6 cells/mL were cultivated further in 20 mL YPD medium in the presence or absence of fluconazole. Cell growth was monitored by measuring OD 600 , and cells were harvested after the population had doubled 10 times and processed for NGS analysis.
Fluconazole susceptibility test. Antifungal susceptibility testing of the strains included in this study was performed by using the broth microdilution method described in the CLSI document M27-A3 67 in a 96-well plate format.
Fluconazole concentrations tested ranged from 64 to 0.25 μg/mL. RPMI 1640 medium (Thermo Fisher Scientific, 21875109) was prepared according to the CLSI document. The medium was buffered with 0.165 M morpholinepropanesulfonic acid (MOPS), and its pH was adjusted using NaOH and HCl. Cell inoculum was 1 × 10 3 cells per well. Plates were incubated without shaking at 30°C for 24 h. The content of each well was carefully resuspended by pipetting up and down before OD 600 was measured using the Spark multimode microplate reader (TECAN, Switzerland).
Preparation of C. albicans plasma membrane. Highly purified plasma membrane was isolated as previously described for S. cerevisiae with minor modifications. C. albicans yeast cells were harvested by centrifugation (5000g, 5 min), then washed twice with 0.4 M sucrose (Sigma, V900116-500G) in buffer A (25 mM imidazole, pH 7.0 adjusted by HCl). Enough 0.4 M sucrose in buffer A and glass beads were added to the cell pellet and the mix was vortexed for 1 min 8-10 times, with 2 min on ice between rounds. Then the mix was centrifuged at 530g (2500 rpm in Sorvall SS34 rotor) for 20 min to pellet unbroken cells and glass beads, and the supernatant was centrifuged at 22,000g (16,000 rpm in Sorvall SS34 rotor) for 30 min to obtain a pellet that includes the plasma membranes and mitochondria. The pellet was resuspended in buffer A by gentle vortexing for 30 s and the resuspended membranes were loaded onto discontinuous sucrose gradients (overlaying three 4 mL layers of 2.25, 1.65, and 1.1 M sucrose in buffer A in a 14 × 89 mm Beckman ultraclear tube) and centrifuged either overnight (14 h) at 80,000g (22,000 rpm) or for 6 h at 284,000g (40,000 rpm) in the Beckmann SW41 or SW40Ti rotor. The membrane band at the 2.25/1.65 M sucrose interface is essentially pure plasma membrane, which was collected at these interfaces from the top of the gradient with a Pasteur pipette. The collected membrane was diluted four times with buffer A and pelleted at 30,000g (18,000 rpm Beckman 50 Ti rotor) for 40 min. The plasma membrane pellet was suspended in 80% methanol and stored at −80°C for sterol analysis.
Sterol analysis. Sterols were extracted using pre-chilled 80% (v/v) methanol (Sigma, 34860-1L-R) according to the method of Yuan et al. 68 . C. albicans strains were cultured in YPD medium with or without fluconazole for 24 h, and the PBS washed cells were harvested by centrifugation. Add pre-chilled 80% (v/v) methanol and glass beads to the cell pellets. Vortex 1 min, then keep on ice for 2 min. Repeat this step 8 or 10 more times. Centrifuge at 14,000g for 40 min using a refrigerated centrifuge at 4°C to obtain the supernatant. The extracts were vacuum-dried and stored at −80°C. The dried samples were solubilized in methanol prior to analysis by LC-MS. Equal amounts of cells were processed for protein extraction, and protein concentration was measured by Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific, 23227) to conduct normalization.
LC-MS analysis was performed with a UPLC system, which was coupled to a Q-Exactive orbitrap mass spectrometer (Thermo Fisher, CA) equipped with an atmospheric pressure chemical ionization (APCI) probe. Sterol extracts were separated by a Kinetex® 100 × 2.1 mm 2.6 μm column (Phenomenex). A binary solvent system was used, in which mobile phase A consisted of 100% H 2 O (0.1% FA), 10 mM ammonium acetate, and mobile phase B of 100% ACN (0.1% FA). A 10-min gradient with a flow rate of 300 μL/min was used. Column chamber and sample tray were held at 45 and 10°C, respectively. Data with mass ranges of m/z 70-150 was acquired at a positive ion mode with data dependent MSMS acquisition. The full scan and fragment spectra were collected with a resolution of 70,000 and 17,500, respectively. The source parameters are as follows: discharge current: 6 μA; capillary temperature: 350°C; heater temperature: 300°C; sheath gas flow rate: 35 Arb; auxiliary gas flow rate: 15 Arb. Data analysis was performed by the software Xcalibur (Thermo Fisher, CA). Compounds were identified based on their retention times and the masses detected in the liquid chromatography-mass spectrometry system. Lipid analysis. Lipids were extracted according to the method of Bligh and Dyer 69 . C. albicans strains were cultured in YPD medium with or without fluconazole for 24 h, and the PBS washed cells were harvested by centrifugation. The cell pellets were lysed in PBS by bead-beating mechanical disruption at 4°C. The supernatants were then extracted with chloroform (AMRESCO, 0757-500ML)/methanol (2:1) at a final ratio of 20% (v/v). Vortex 30 s, then keep for 2 min. Repeat this step 5 more times. Centrifuge at 3000g for 20 min using a refrigerated centrifuge at 4°C to obtain the supernatant. The extracts were evaporated to dryness under N 2 at room temperature and stored at −80°C. The dried samples were solubilized in dichloromethane (Sigma, 650463-1L) methanol (2:1) before analysis by LC-MS. Equal amounts of cells were processed for protein extraction and protein concentration was measured by BCA assay to conduct normalization.
LC-MS analysis was performed with a UPLC system was coupled to a Q-Exactive orbitrap mass spectrometer (Thermo Fisher, CA) equipped with a heated electrospray ionization (HESI) probe. Lipid extracts were separated by a CORTECS C18 100 × 2.1 mm 1.9 μm column (Waters). A binary solvent system was used, in which mobile phase A consisted of ACN:H 2 O (60:40), 10 mM ammonium acetate, and mobile phase B of IPA:ACN (90:10), 10 mM ammonium acetate. A 35-min gradient with a flow rate of 220 μL/min was used. Column chamber and sample tray were held at 45 and 10°C, respectively. Data with mass ranges of m/z 240-2000 and m/z 200-2000 was acquired at positive ion mode and negative ion mode with data dependent MSMS acquisition. The full scan and fragment spectra were collected with a resolution of 70,000 and 17,500, respectively. The source parameters are as follows: spray voltage: 3000 V; capillary temperature: 320°C; heater temperature: 300°C; sheath gas flow rate: 35 Arb; auxiliary gas flow rate: 10 Arb. Data analysis and lipid identification were performed by the software Lipidsearch 4.0 (Thermo Fisher, CA). Compounds were identified based on their retention times and the masses detected in the liquid chromatography-mass spectrometry system.