Step-wise acquisition of resistance during adaptation to fluconazole

We recently identified a lineage (AMS4058) in an evolution experiment in which the drug sensitive progenitor strain SC5314 acquired FLC resistance after 10 passages in 1 μg/mL FLC, resulting in an increase in MIC₅₀ from 0.5 μg/mL to >256 μg/mL FLC [49]. The AMS4058 lineage also had reduced filamentation in vitro and decreased virulence in a mouse model of disseminated candidiasis (S2A and S2B Fig). To understand the dynamics of adaptation in lineage AMS4058, we measured the MIC₅₀ at each of the 10 passages. The MIC₅₀ increased 2-fold after passage 1 and again after passage 3; and increased dramatically to >256 μg/mL FLC after passage 4 (Fig 1A). We also measured FLC tolerance after each passage (supra-MIC growth, SMG, see Methods) [21,28]. SMG increased dramatically after the first passage and steadily decreased in passages 2 and 3 (Fig 1B). We hypothesized that variants conferring large increases in tolerance were dominant early in the evolution experiment (passages 1 to 3), whereas variants conferring resistance arose around passage 4 and continued to increase in frequency in all subsequent passages.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Step-wise evolution of drug resistance via CNV and LOH.

(A) MIC₅₀, defined as the drug concentration at which growth is inhibited by >50%, is plotted on the y-axis for each passage of the evolution experiment. (B) SMG, defined as the average growth in drug concentrations greater than the MIC₅₀, is plotted on the y-axis for each passage of the evolution experiment. (C) Read depth from whole genome sequencing of the entire population for each passage of the evolution experiment is plotted according to genomic position and normalized to the average read depth for the entire genome. (D) Median copy number, calculated from read depth relative to the entire genome, for whole chromosomes with variation in copy number (Chr3 and Chr5) is plotted on the y-axis for the entire population at each passage of the evolution experiment. (E) Median copy number of the central amplified portion of Chr4, calculated from read depth in the entire passaged population relative to the rest of the genome, is plotted on the y-axis at each passage of the evolution experiment and connected by line segments. (F) Relative copy number for the entire population at passage 9 at the region containing a CNV, calculated from the relative average read depth, is plotted on the y-axis according to the position on Chr4 on the x-axis. Each point represents the average read depth of a 500 bp window. Dashed lines show the location of repeat sequences in the C. albicans genome [52]. (G) OD₆₀₀ values for liquid culture growth assays are plotted over time for the wild-type progenitor (black), a single colony containing the Chr4 amplification without LOH at KSR1 (P4.4 single colony, light blue), and a single colony containing the Chr4 amplification with the LOH at KSR1 (P4.5 single colony, dark blue). Growth in rich media, 1 μg/mL FLC, and 256 μg/mL FLC are shown. Error bars are standard errors for three replicates. Points are averages. A table to the right shows average MIC₅₀ and SMG values for the listed strains. (H) Allele frequencies for 2 single nucleotide variants at KSR1 amino acid positions 189 and 192 are plotted for each passage of the evolution experiment. Reference “A” alleles are shown in cyan, and reference “B” alleles are shown in magenta. (I) Biallelic single nucleotide polymorphism (SNP) frequencies are plotted across the KSR1 locus on chromosome R for two single colonies isolated from passage 4, with MIC₅₀ = 1 (P4.4, top) and MIC₅₀ >256 μg/mL FLC (P4.5, bottom). The absence of biallelic SNPs in P4.5 indicates a loss of heterozygosity highlighted by the pink-shaded region. (J) OD₆₀₀ values for liquid culture growth assays are plotted over time for the wild-type progenitor (black), the P10 evolved population (red), and a strain engineered to contain the LOH at KSR1 (pink). Growth in rich media with 0 μg/mL, 1 μg/mL, or 256 μg/mL FLC are shown. Error bars are standard errors for three replicates. Points are averages. A table to the right shows MIC₅₀ and SMG values for the listed strains.

https://doi.org/10.1371/journal.ppat.1012497.g001

Aneuploidy and copy number variation arise early and yield decreased susceptibility and increased tolerance

To identify the genetic changes associated with decreased drug susceptibility and increased tolerance, we performed population-level whole genome sequencing of each passage of the evolution experiment. We first estimated the population-average copy number of each chromosome by examining sequencing read depth across the genome for all 10 passages. Increases in read depth indicative of copy number variants (CNVs) were present in the population from passage 2 to passage 10 (Fig 1C–1E). The median copy number of Chr3 was ~2.18, 2.27, and 2.15 for passages 1, 2, and 3, respectively (Fig 1D), suggesting that ~15–30% of the population was trisomic for Chr3 at these early passages with high tolerance (see Methods). Chr5 also rose to a frequency of ~2.17 by passage 2 (Fig 1D). Chr3 contains several genes encoding efflux pumps and their regulators, such as Cdr1, Cdr2, and Mrr1; and Chr5 contains genes encoding Erg11 (the azole drug target) and efflux pump regulator Tac1. Additional copies of these genes have the potential to contribute to decreased susceptibility and increased tolerance observed in the early populations of the evolution experiment.

In addition to evidence of aneuploidy, we detected a segmental amplification of Chr4 (Fig 1C and 1E), consisting of a stair-step structure with two flanking regions (~20–25 kb, A21 Chr4 coordinates ~532000 to ~553000, and ~681000 to ~704500) surrounding a central region with the highest copy number (~100 kb, A21 Chr4 coordinates ~553000 to ~674000) (Fig 1F). Each amplified region was flanked by inverted repeat sequences, consistent with segmental amplifications as previously described (Fig 1F, [52]). Median read depth in the central region increased from a population average relative copy number of 2 in passage 1 to ~3.5 by passage 2, ~5 by passage 3, and then gradually rose to ~6 copies by passage 10 (Fig 1E).

To identify individual genotypes within the passaged populations, we isolated and sequenced 5 single colonies each from passages 2, 3, and 4. Passage 2 included one colony with wild type chromosome copy numbers (euploid), one colony with an LOH of ChrR and duplication of the B homolog, resulting in a ChrR BB genotype, two colonies with monosomy of the right end of ChrR, and one colony with homozygosis of the left arm of Chr4 resulting in an AA genotype (S3A Fig). For each unique genotype from passage 2 we performed growth curve and microbroth dilution assays and found that the colony with ChrR LOH grew faster than the progenitor strain in all concentrations of FLC, with an MIC₅₀ of 2 μg/mL FLC and an SMG of 0.78 (S3B Fig). In passage 3 we found two distinct genotypes: 2 colonies displayed concurrent trisomy of Chr3 and Chr6 along with LOH encompassing the entirety of ChrR resulting in a ChrR AA genotype; and 3 colonies displayed the Chr4 segmental amplification described above (S3C Fig). Both genotypes from passage 3 also grew better than the progenitor in FLC in growth curve assays and had a slightly elevated MIC₅₀ of 1 μg/mL FLC and SMG of 0.25 and 0.3 (S3D Fig). Thus, early in the evolution experiment, different copy number variants arose and conferred a 2- to 4-fold increase in MIC₅₀ and increased tolerance.

The amplified region of Chr4 does not contain genes known to influence drug susceptibility upon overexpression. This region includes ~88 genes, including NCP1, which encodes the Erg11p reductase (S4A Fig). We tested the hypothesis that overexpression of Ncp1p is sufficient to promote growth in 1 μg/mL FLC by engineering strains with NCP1 under the control of a tet-off promoter system inserted at one of the two copies of the native locus. Without the addition of doxycycline, these strains overexpressed NCP1 at ~8-fold the level of the wild-type strain; with the addition of doxycycline, NCP1 was expressed at 0.5-fold relative to its expression in wild-type cells (S4B Fig). This level of NCP1 overexpression closely matched that of strains containing the Chr4 CNV in FLC, which was also ~8-fold relative to the wild-type strain (S4C Fig). Overexpression of NCP1 in these engineered strains resulted in a 2-fold increase in MIC₅₀ that was abolished with the addition of doxycycline (S4D Fig). Because there is allelic variation at the NCP1 locus in SC5314, we overexpressed the A and B alleles independently; both NCP1 alleles resulted in the same 2-fold increase in MIC₅₀ (S1 Table). Thus, overexpression of NCP1 is sufficient to increase growth in low drug concentrations and provides at least one explanation for the increased MIC₅₀ in the single colonies carrying the Chr4 CNV. However, overexpression of NCP1 cannot explain the MIC₅₀ of >256 μg/mL FLC displayed by passages 4–10 of the evolution experiment.

LOH at KSR1 produces a resistance phenotype

To identify the genetic changes responsible for the resistance seen in passages 4–10, we examined the genomes of 5 single colonies from passage 4 and measured their growth across a range of FLC concentrations. All 5 colonies contained the Chr4 CNV (S3E Fig); however, only 1 of the 5 colonies, isolate P4.5, grew significantly faster than the progenitor in 256 μg/mL FLC and had an MIC₅₀ of >256 μg/mL FLC (S1 Table and Fig 1G). Interestingly, P4.5 contained a short (711 bp) LOH of the ChrR B homolog between coordinates 1604191 to 1604902 that was not present in the other colonies. Population-level sequencing of each passage showed that allele ratios in this region were biased for allele A in passages 1–3 but switched to a bias for allele B in passage 4 and steadily rose to 100% frequency of the B allele by passage 9 (Fig 1H). This ~700 bp LOH (which we will now call LOH1) encompassed ~⅔ of KSR1 (CR_07380C) which encodes 3-ketosphinganine reductase (Fig 1I) and catalyzes the second step in sphingolipid biosynthesis [39].

To directly test the effect of the 711 bp LOH at KSR1 (KSR1 LOH1) on drug susceptibility, we replaced this portion of the KSR1A allele with KSR1B, reconstructing the evolved LOH in the progenitor background (SC5314) (without the Chr4 CNV). This strain, (KSR1 LOH1) grew faster than the progenitor at FLC concentrations from 1 to 256 μg/mL, with an MIC₅₀ of 8 μg/mL FLC and an SMG of 0.86 (Fig 1J). Thus, KSR1 LOH1 alone is sufficient to cause a 16-fold decrease in susceptibility and a high tolerance phenotype (~8-fold higher than the progenitor).

Next, we tested the hypothesis that overexpression of NCP1 and the KSR1 LOH1 in combination could explain the evolved resistance phenotype. A strain engineered to carry both the tetO-NCP1 overexpression construct and the KSR1 LOH1 genotype (tetO-NCP1, KSR1 LOH1) grew faster than either single mutant alone in low and high FLC. However, the strain with tetO-NCP1 and KSR1 LOH1 had a final MIC₅₀ of 4 and SMG of 0.66 (S5 Fig), which is not sufficient to reproduce the MIC₅₀ of single colony P4.5, which carried the ~150kb Chr4 CNV and the KSR1 LOH1. This implies that one or more additional genes within the Chr4 CNV might be necessary, in combination with the KSR1 LOH1, to account for the high MIC₅₀ of the populations at passages 4–10 and of single colony P4.5. Additionally, amplification of the Chr4 CNV coincided with LOH of Chr4 sequences that flank the CNV (S2C Fig). Therefore, we cannot rule out the possibility that LOH of other genes on Chr4 could contribute to high resistance in strain P4.5 as well.

To ask if Ksr1p is necessary for drug resistance, we tried to engineer null strains lacking both copies of KSR1. We obtained heterozygous mutants by deleting either the A or B allele individually. Strains lacking the KSR1 B allele had a wild-type MIC₅₀ of 0.5 μg/mL FLC, while strains lacking the KSR1 A allele had an MIC₅₀ of 2 μg/ml FLC, a 4-fold increase relative to the parent strain (S6A Fig). Despite several attempts in multiple labs, the only two ksr1Δ/Δ transformants recovered had undergone large LOHs encompassing most of ChrR (S6B Fig), which confounded the interpretation of phenotypes associated with the complete loss of Ksr1p. In Saccharomyces cerevisiae, the KSR1 ortholog (TSC10) is essential [54]; however, in C. albicans, null mutants of KSR1 (strains SC90, SC91) were previously described in strain BWP17, an auxotrophic derivative of SC5314 [39]. Our whole genome sequencing of these original BWP17 ksr1Δ/Δ null strains revealed that both mutants had also undergone LOH encompassing most of ChrR (S6B Fig). Interestingly, two of the four ksr1Δ/Δ null strains were homozygous for ChrR AA (blue) and the other two were homozygous for ChrR BB (pink), but in all four strains the region flanking the KSR1 locus was homozygous for the A allele (AA) (S6C Fig). Importantly, other manipulations of KSR1 usually were not accompanied by large LOHs (see below), suggesting that KSR1 might be essential in the SC5314 background when ChrR maintains normal heterozygosity.

KSR1 LOH1 results in reduced sphingolipids, lower intracellular FLC, higher drug efflux pump activity, and decreased toxic sterol content

Ksr1p catalyzes an early step in sphingolipid biosynthesis (S1 Fig) and changes in its function may alter the cellular sphingolipid composition [39,55,56]. Consistent with this, ksr1Δ/Δ strains (SC90 and SC91, containing large ChrR LOH regions) had lower levels of inositol phosphorylceramides and were sensitized to inhibitors of subsequent steps of sphingolipid biosynthesis [39]. To determine the effect of the evolved KSR1 LOH1 we compared the lipid content of the progenitor (SC5314) and KSR1 LOH1 (engineered into SC5314), with and without FLC treatment, using mass spectrometry. In the absence of FLC, the abundance of all sphingolipid species was reduced in the KSR1 LOH1 strain relative to the progenitor (Fig 2A). In the progenitor, upon FLC exposure (1 and 10 μg/mL) intermediates of the sphingolipid pathway increased and glucosylceramides decreased in abundance (Fig 2A). By contrast, the KSR1 LOH1 strain had consistently low levels of both phytosphingolipids and glucosylceramides. In addition, exposure of the KSR1 LOH1 strain to 10 μg/mL FLC resulted in increased sphingosine (Sph), phosphorylated dihydrosphingosine (dhSph-1-p) and phosphorylated phytosphingosine (PhytoSph-1-p) relative to no drug conditions, in two of the three replicates (Fig 2A). No such change in sphingolipid species was seen in the progenitor. Together, these results indicate that in general, sphingolipid species are decreased in the KSR1 LOH1 strain relative to the progenitor strain, and that the two strains respond differently to FLC.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. LOH at KSR1 affects cellular lipid profiles, intracellular FLC concentrations, drug efflux, and sterol levels.

(A) A heatmap shows abundance of sphingolipid species relative to inorganic phosphate present in each strain and condition relative to the mean of all samples for each respective lipid species. Heatmap colors indicate row-wise z-scores, red being high and blue being low. Rows are lipid species and columns are individual samples, both of which are clustered by k-means clustering as shown by the dendrograms on the top and left. Columns include 3 replicates each of the wild type progenitor strain (black) grown in 0 μg/mL (white), 1 μg/mL (grey), and 10 μg/mL FLC (black), and the KSR1 LOH1 engineered strain (purple) in the same conditions. Two outlier samples were removed (WT in 10 μg/mL FLC and KSR1 LOH1 in 1 μg/mL FLC, see Methods). Wild type samples cluster according to condition, indicating consistent changes in sphingolipid species as FLC concentrations are increased. KSR1 LOH1 samples do not consistently cluster by condition, indicating only subtle changes in sphingolipid species as FLC concentration increases. An abbreviated sphingolipid biosynthesis pathway is included for reference (see full pathway at S1 Fig). (B) Mean intracellular Rhodamine 6G (R6G) fluorescence, as measured by flow cytometry, is plotted with both points and bars for each strain listed after 4 hours of incubation with R6G. Each point represents one replicate, bars are means and whiskers are quartiles. (**Indicates a paired t-test p-value of < 0.01) (C) Extracellular R6G fluorescence is plotted on the y-axis over time (on the x-axis) after addition of glucose, for the progenitor strain (black) and engineered KSR1 LOH1 (purple). Points are means and error bars are standard errors of the mean across 3 replicates. (D) As in (B), mean intracellular Cy5 (FLC-Cy5) fluorescence is plotted for each strain listed after incubation with FLC-Cy5 for 4 hours, with or without the addition of beauvericin. (E) Ergosterol content relative to inorganic phosphate as measured by GC-MS are shown for the progenitor strain and KSR1 LOH1 strain after exposure to 0, 1, and 10 μg/mL FLC for 4 hours. (F) Abundance of the toxic dienol (14α-methylergosta-8,24(28)-dien-3β,6α-diol, peak at 24.3 min retention time) relative to ergosterol abundance is shown for the same samples shown in (E). (D and E) **Indicates a paired t-test p-value of < 0.05, *Indicates a p-value of < 0.1. Each point represents one replicate, bars are means and whiskers are quartiles for 3 biological replicates.

https://doi.org/10.1371/journal.ppat.1012497.g002

A reduction in sphingolipid species could significantly alter membrane structure, membrane fluidity, and the formation of lipid rafts, which help properly position membrane-associated proteins such as the drug efflux pump Cdr1p [25]. In addition, PhytoSph-1-P is thought to be a signaling molecule that can lead to the upregulation of drug efflux pumps such as Cdr1p and Cdr2p and a reduction in intracellular FLC [34]. Thus, changes in the sphingolipid composition and structure of the cell membrane might influence the activity of drug efflux pumps.

To evaluate the effect of KSR1 LOH1 on efflux pump activity, we first monitored intracellular levels of Rhodamine 6G (R6G), a fluorescent molecule that is actively effluxed by ABC transporters [57], by flow cytometry after incubation for 4 hours. KSR1 LOH1 intracellular R6G levels were ~3-fold lower than in the progenitor strain (Fig 2B, paired t-test p-value = 0.009). To measure the dynamics of active drug efflux, we performed a time-course in which cells were preloaded with R6G, and fluorescence of the supernatant was measured over time after the addition of glucose. The KSR1 LOH1 strain had a ~6-fold increase in the rate of efflux relative to the progenitor strain after growth in 0, 1, and 10 μg/mL FLC (Fig 2C). In addition, using RT-qPCR, we found that, relative to the progenitor, CDR1 was upregulated ~1.5 fold in the absence of FLC, and ~2 fold in 1 μg/mL FLC (S7 Fig). While this increase in expression likely contributes to the higher rate of efflux, the much more dramatic increase in efflux (~6-fold increase in rate, Fig 2C) suggests that additional mechanisms influence efflux activity. We next tested whether this increased efflux affected intracellular drug concentrations of cells after incubation for 4 hours with a fluorescently labeled FLC probe (FLC-Cy5, [58]), in the presence and absence of the ABC efflux pump inhibitor beauvericin. The KSR1 LOH1 strain had significantly lower intracellular levels of FLC-Cy5 than the progenitor strain at 4 hours (Fig 2D, paired t-test p-value = 0.004). However, intracellular levels of FLC-Cy5 remained significantly lower than the progenitor strain in the presence of the efflux pump inhibitor beauvericin (Fig 2D), suggesting that efflux by ABC transporters alone cannot account for the differences in intracellular FLC levels. Together, these results indicate that cells with the KSR1 LOH1 have higher rates of drug efflux and accumulate lower levels of intracellular FLC than the progenitor strain.

Increased activity of drug efflux pumps, likely resulting from the changes to the sphingolipid components of the membrane, can elevate cellular ergosterol levels even in the absence of FLC [22,23]. Therefore, we assayed sterol levels in progenitor and KSR1 LOH1 cells in the presence of 0, 1, and 10 μg/mL FLC to determine the effect of the LOH on ergosterol and other sterols. We found that in the absence of drug, the KSR1 LOH1 strain had a significantly higher ergosterol abundance than the progenitor (Fig 2E, two-sided t-test p-value = 0.099). The KSR1 LOH1 strain also showed increased sensitivity to Amphotericin B, which targets membrane ergosterol (S8 Fig). In the presence of both 1 μg/mL and 10 μg/mL FLC, however, ergosterol abundance was significantly decreased to similar levels in both the progenitor and KSR1 LOH1 strains (Fig 2E, p-values left to right: 0.028, 0.027, 0.009, 0.015). This suggests that reduced intracellular levels of FLC are not associated with increased ergosterol in the KSR1 LOH1 strain. Rather, elevated efflux pump activity is indicative of changes to membrane homeostasis that make cells less sensitive to FLC. In agreement with this, the abundance of a toxic dienol (14α-methylergosta-8,24(28)-dien-3β,6α-diol) known to inhibit growth [59] relative to ergosterol levels was significantly lower in KSR1 LOH1 cells compared to progenitor cells in the presence of 10 μg/mL FLC (Fig 2F, two-sided t-test p-value = 0.0053). Together these results indicate that the KSR1 LOH1 results in a reduction in sphingolipid species and changes to membrane homeostasis, which are accompanied by increased drug efflux pump activity and lower levels of the toxic dienol in the presence of FLC.

Homozygosis of arginine at Ksr1p aa189 is the major driver of decreased drug susceptibility

To examine the effects of specific amino acid(s) in Ksr1p that contribute to the drug susceptibility phenotype, we compared KSR1 LOH1 to three other independently evolved SC5314-strains that acquired KSR1 LOH events during adaptation to 1 μg/mL FLC (Evolved LOH2, Evolved LOH3, Evolved LOH4, see Methods). These three evolved LOH events ranged in size from ~20 bp to ~2,000 bp of ChrR and each included a portion of the KSR1 locus (Figs 3A and S9A and S1 Table). All three evolved strains had 4-fold increases in MIC₅₀ and increased SMG relative to the progenitor strain (Fig 3B). They also were defective in filamentation and had decreased virulence in a Galleria mellonella model of systemic fungal infection (S9 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. LOH resulting in 189R/R produces a strong decrease in drug susceptibility.

(A) A schematic of the KSR1 locus including heterozygous positions (black dots) that are either synonymous (S), missense (M), or nonsense (N, *) in the B allele relative to the A allele. A allele is shown in blue and B allele is shown in pink. Allele frequencies are plotted across the KSR1 locus and a larger flanking region on chromosome R for the wild type progenitor (top) and the three evolved isolates (Evolved LOH2, Evolved LOH3, and Evolved LOH4). The absence of biallelic SNVs in the evolved isolates, or allele frequencies of 1, indicates a loss of heterozygosity in that region highlighted by the pink shaded region. Note that the chromosome coordinates are inverted on the map because the KSR1 coding sequence is on the ‘Crick’ strand of the chromosome. (B) OD₆₀₀ values for liquid culture growth assays are plotted over time for the wild type progenitor (SC5314, black), and the three evolved strains with LOH including some portion of the KSR1 locus (Evolved LOH2, Evolved LOH3, and Evolved LOH4). Growth in rich media with 0 μg/mL, 1 μg/mL, and 256 μg/mL FLC are shown. Error bars are standard errors for three replicates. MIC₅₀ and SMG values are shown in a table to the right. (C) Schematics show portions of the KSR1 coding region that underwent LOH in the four evolved LOH events. Dotted lines demarcate the single nonsynonymous SNV that was homozygous in all four LOH isolates, a nucleotide coding for glutamine in allele A and arginine in allele B at amino acid position 189. (D) Liquid growth assays, plotted as in (B), are shown for the progenitor strain (black), the strain engineered to contain the KSR1 LOH1 (purple), and the KSR1^189R/R engineered strain (pink).

https://doi.org/10.1371/journal.ppat.1012497.g003

RNA-sequencing of these three evolved strains when exposed to 0, 2, or 64 μg/mL FLC concentrations was analyzed for principal components. The first principal component, explaining the majority of the variance in the data (83%), separated samples by FLC treatment (S10A Fig). The second principal component, explaining 5% of the variance in the data, separated strains that were exposed to FLC by genotype (S10B Fig). This indicates that both wild type and evolved strains undergo major transcriptional changes upon exposure to FLC, and that, in the absence of FLC, transcriptional differences between progenitor and evolved strains are minimal. Expression levels of genes annotated as related to sphingolipid biosynthesis showed a similar trend, with samples clustering according to FLC exposure first, and then clustering weakly by genotype (S10C Fig). Although 5 genes catalyzing steps in sphingolipid biosynthesis were differentially expressed in all three evolved strains relative to the progenitor strain, no strong correlation between transcriptional changes and changes in corresponding sphingolipid metabolites was evident (S10D Fig), likely because most upregulated genes were downstream of KSR1 in the biosynthesis pathway (S1 Fig). Because the KSR1 LOH1 strain had high levels of glucose-dependent efflux in the R6G assay (Fig 2C), we also examined the expression of ABC drug efflux pumps. Efflux pumps CDR1 and CDR2 were not significantly upregulated in the three evolved strains (S10E Fig). SNQ2, which is annotated as a multidrug transporter in S. cerevisiae [60], was upregulated in all three strains (S10E Fig). Taken together, these data indicate that LOH affecting Ksr1p has a small effect on the global transcriptome and likely impacts the activity and/or localization of drug efflux pumps post-transcriptionally, and that the progenitor and evolved strains respond differently to FLC exposure.

In SC5314, the KSR1 protein coding region harbors 6 synonymous and 4 nonsynonymous heterozygous nucleotides. The LOH events in all four FLC-evolved LOH strains became homozygous for the B allele at a single nonsynonymous variant encoding aa189 (Fig 3C). In the wild-type SC5314, allele A encodes glutamine and allele B encodes arginine (KSR1^189Q/R); in all four evolved LOH isolates (LOH1-LOH4) the sequence encodes arginine at aa189 (KSR1^189R/R). To test the effect of aa189R/R alone, we constructed a strain encoding KSR1^189R/R and measured its growth in a range of FLC concentrations. The KSR1^189R/R strain grew better than the progenitor across all FLC concentrations, with growth dynamics similar to those of the KSR1 LOH1 strain and had the same MIC₅₀ and SMG values (Fig 3D). Thus, KSR1^189R/R is a major driver of the decreased drug susceptibility in all four evolved strains (Fig 3D).

Recurrent and precise LOH breakpoint unlinks the effect of 189R/R from the masking effect of the nonsense codon

KSR1 codon 272 in the SC5314 progenitor encodes arginine on the A allele and encodes a nonsense mutation on the B allele (KSR1^272R/*). Notably, all four LOH events in FLC-resistant strains remained heterozygous at this codon (KSR1^272R/*, Fig 3C). The precision of the right LOH breakpoint between aa189 and aa272 in all four evolved strains could indicate strong selective pressure against homozygosis of the nonsense codon at aa272. We asked if a strain carrying the homozygous nonsense allele alone would be viable, and if so, if the nonsense allele would mask the effect of aa189R/R on drug susceptibility. To address these questions, we engineered mutants that were homozygous for the nonsense codon (KSR1²⁷²*^/*) and mutants that were homozygous for the entire B allele (KSR1B/B, which is homozygous for both aa189R/R and aa272*/*) (Fig 4A). We were able to recover both mutants without any additional LOH events (S11 Fig), and found that both KSR1²⁷²*^/* and KSR1B/B had increased MIC₅₀ and SMG values (MIC₅₀ = 4 μg/mL, SMG = 0.86) relative to the progenitor SC5314 (MIC₅₀ = 0.5 μg/mL, SMG = 0.01), but the MIC₅₀ was not elevated to the same degree as KSR1 LOH1 (MIC₅₀ = 8 μg/mL, SMG = 0.86). Furthermore, growth curve analysis across a range of FLC concentrations determined that the average growth of KSR1B/B and KSR1²⁷²*^/* was nearly identical and consistently lower than the growth of KSR1 LOH1 and KSR1^189R/R strains at drug concentrations > 1 μg/mL (Fig 4B, paired t-test between KSR1 LOH1 and KSR1B/B p-values < 0.1). The matched growth of KSR1²⁷²*^/* and KSR1B/B (which includes aa189R/R and aa272*/*), indicates that homozygosis of the nonsense codon affects drug susceptibility itself and masks the effect of 189R/R in strains carrying both 189R/R and 272*/*.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. LOH unlinks 189R/R from a stop codon with masking effects on phenotype.

(A) Schematics of the KSR1 locus in four engineered strains: the strain engineered to contain the KSR1 LOH1 (as in Fig 1J), one that is homozygous for arginine at aa189 only (KSR1^189R/R), one that is completely homozygous for the entire KSR1B allele (KSR1B/B), and one that is homozygous for the nonsense codon to the end of the protein (KSR1²⁷²*^/*). SNP positions are indicated by circles, and the SNP that codes for the nonsense codon in allele B is indicated by an asterisk. A and B alleles are indicated in blue and pink, respectively. (B) Area under the growth curve, calculated from 48 h liquid growth assays, is plotted on the y-axis, relative to the growth condition of various FLC concentrations shown on the x-axis. Each point is the mean of three replicates. Each strain is shown in a different color and connected by line segments. (*Indicates paired t-test values < 0.1, ** indicates paired t-test values < 0.01 for all comparisons in 2ug/mL FLC and higher) (C) Mean intracellular Cy5 fluorescence, as measured by flow cytometry, is plotted with both points and bars for each strain listed, after 4 hours of incubation with FLC-Cy5. Each point represents one replicate, bars are means and whiskers are quartiles. The same data as shown in Fig 2B is shown with the addition of the KSR1B/B strain for comparison, and separate panels are separate experiments (**Indicates a paired t-test p-value of < 0.01). (D) Extracellular R6G fluorescence is plotted on the y-axis over time after the addition of glucose, measured in minutes, on the x-axis. The same data as shown in Fig 2D is shown with the addition of a KSR1^189R/R (pink), KSR1B/B (light blue), and KSR1²⁷²*^/* (dark blue) strains for comparison. (E) Spot plates showing serial dilutions of strain across the x-axis at concentrations of 10⁷, 10⁶, 10⁵, 10⁴, and 10³ cells/mL in rich media (YPAD), and 0.02% sodium dodecyl sulfate (SDS). One representative replicate of 3 performed is shown. (F) A heatmap showing levels of sphingolipid species present in each strain and condition relative to the mean among all samples for that lipid species. Heatmap colors indicate row-wise z-scores, red being high and blue being low. Rows are lipid species, clustered by k-means clustering as shown by the dendrograms on the left, and columns are individual samples, which are not clustered. Columns include 3 replicates each of the wild-type progenitor strain (black) grown in 0 μg/mL (white), 1 μg/mL (grey), and 10 μg/mL FLC (black), and the KSR1 LOH1 engineered strain (purple), the KSR1B/B strain (light blue) and KSR1²⁷²*^/* (dark blue) in the same conditions.

https://doi.org/10.1371/journal.ppat.1012497.g004

We next asked if homozygosis of the nonsense codon 272*/* and homozygosis of 189R/R affect drug susceptibility via similar mechanisms by measuring intracellular levels of FLC at 4 hours after drug exposure. We found that KSR1 LOH1, KSR1^189R/R, and KSR1B/B strains accumulated less intracellular FLC relative to the progenitor strain (Fig 4C, paired t-test p-value for KSR1B/B = 0.004, for KSR1^189R/R = 0.004). Furthermore, KSR1 LOH1, KSR1^189R/R, KSR1²⁷²*^/*, and KSR1B/B all had increased active efflux of R6G when exposed to 0, 1, or 10 μg/mL FLC (Fig 4D) and increased susceptibility to Amphotericin B (S8 Fig). All four mutants also had a heightened sensitivity to the detergent SDS, which destabilizes cell membranes (Fig 4E). Interestingly, the strains with a homozygous nonsense codon (KSR1B/B and KSR1²⁷²*^/*) had a stronger growth defect in the presence of SDS relative to the KSR1 LOH1 or KSR1^189R/R strains (Fig 4E). Thus, all mutants had similar intracellular FLC levels and efflux levels, but those with a homozygous nonsense codon were more sensitive to SDS, suggestive of a more severe membrane defect.

We next asked about differences in the sphingolipid composition of KSR1 LOH1 relative to strains homozygous for the nonsense codon. By comparing lipidomic data for the KSR1 LOH1, KSR1B/B, and KSR1²⁷²*^/* strains and the progenitor, we found that strains homozygous for the nonsense mutation had lower levels of most sphingolipid species relative to both the progenitor and the KSR1 LOH1 strain (Fig 4F, purple box). In addition, strains homozygous for the nonsense mutation did not have the increase in dhSph-1-p or PhytoSph-1-p that was seen in the KSR1 LOH1 strain relative to the progenitor at 10 μg/mL FLC (Fig 4F). Together, these results suggest that both KSR1 LOH1 and strains homozygous for the nonsense codon have decreased drug susceptibility related to increased drug efflux pump activity likely due to changes in their sphingolipid composition, and that changes in sphingolipid levels are more severe in strains homozygous for the nonsense codon.

The SDS sensitivity of the strains prompted us to examine their vacuole morphology, which can be an indication of membrane stress [37]. Indeed, using FM4-64 to stain vacuolar membranes, we observed some vacuolar abnormality: in FLC, both KSR1 LOH1 and KSR1B/B strains had smaller vacuolar area than the progenitor strain (S12A and S12B Fig), yet cells were similar in size and morphology to the progenitor strain (S12A Fig).

Lipid droplets affect cellular stress responses by sequestering excess toxic lipids and sterols and mediating the release and transport of lipids from the ER to the plasma membrane and other cellular organelles [61]. Using BODIPY staining to visualize lipid droplets, we found that the KSR1 LOH1 strain had, on average, a higher, but not significantly different, number of lipid droplets per cell as compared to the progenitor, in rich media. Conversely, the KSR1B/B strain had, on average, a significantly lower number of BODIPY stained lipid droplets, which was similar to the decrease in the number of lipid droplets seen when wild-type cells were treated with myriocin (S13A and S13B Fig). Since myriocin inhibits the step in sphingolipid biosynthesis immediately preceding the reaction catalyzed by KSR1 [35,36], this indicates that KSR1B/B shares a phenotype characteristic of sphingolipid pathway inhibition while the KSR1 LOH1 does not. Together, these results indicate that KSR1 LOH1 as well as strains with homozygous nonsense codon affect lipid homeostasis, membrane stress, and the formation of lipid droplets, but that these two types of mutations differ in their effect on lipid droplet properties.

Nonsense codon at aa272 results in mislocalization of Ksr1p

The KSR1 nonsense codon is predicted to truncate the protein immediately upstream of a putative membrane-anchoring domain encoded by KSR1A in C. albicans and in TSC10, the S. cerevisiae ortholog (Fig 5A, [39,62,63]). To ask if this C-terminal tail affects Ksr1p localization, we fused an mNeon-Green fluorescent protein to the N-terminus of either KSR1A or KSR1B in the SC5314 background. As expected based on the localization of Tsc10p in S. cerevisiae, mNeonGreen-Ksr1A co-localized with an ER stain (Fig 5B, [64]). However, mNeonGreen-Ksr1B labeling was diffuse throughout the cytoplasm (Fig 5B). We posit that the failure of the truncated Ksr1Bp to localize to the ER likely prevents efficient contact of Ksr1p with its substrate, which is in the ER, and thus effectively reduces Ksr1Bp activity. The reduction in activity due to mislocalization appears more severe than that resulting from the homozygosis of arginine at aa189, based on the further reduction in sphingolipid species abundance in strains with homozygous nonsense alleles (Fig 4F). However, rather than further increasing drug efflux pump activity, efflux in these strains remained steady (Fig 4C and 4D), and sensitivity to SDS was increased (Fig 4E). Together, these data suggest that cumulative positive and negative fitness trade-offs result in a ‘sweet spot’ that is occupied by the evolved strains with KSR1^189R/R but not in those harboring homozygous nonsense codons (Fig 5C). This underscores the observation that, at the end of the evolution experiments, all four LOH events homozygosed KSR1 aa189 but not aa272.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Nonsense allele results in mislocalization of Ksr1p.

(A) Ribbon models produced by Alphafold [62] predictions are shown for C. albicans Ksr1pA and Ksr1pB. The early stop codon in Ksr1pB (red star) truncates the protein ahead of a putative membrane-anchoring domain positioned on the right side of the model. Models are colored according to Alphafold confidence scores, in which blue is high and yellow to red are low. (B) Fluorescence microscopy images for engineered strains containing KSR1A (top row) or KSR1B (bottom row) tagged with NeonGreen. From left to right the columns show DIC, Neon green fluorescence, the endoplasmic reticulum-specific probe 2 [64], and a merged image of mNeonGreen and ER probe. (C) A model reflecting the levels of sphingolipid abundance, rate of active efflux, and membrane sensitivity together contributing to a theoretical fitness landscape. The regions on this landscape occupied by a completely wild-type Ksr1p (left), the evolved LOH strains that bear alleles homozygous for KSR1^189R/R (middle), and strains with a homozygous nonsense mutation resulting in mislocalization (right) are shown. The evolved LOHs occupy a ‘sweet spot’ with increased active efflux, intermediate sphingolipid levels, and membrane sensitivity leading to the lowest drug susceptibility and highest tolerance.

https://doi.org/10.1371/journal.ppat.1012497.g005

Ethics statement

The mouse studies were approved by the Institutional Animal Care and Use Committee at the Lundquist Institute for Biomedical Innovation at Harbor-UCLA Medical Center.

Microbroth dilution assays (MIC₅₀ and SMG) and growth curve assays

Strains were grown from frozen glycerol stocks overnight in 3 mL of YPAD containing 2% dextrose (20 g/L Bactopeptone, 10 g/L yeast extract, 0.04 g/L adenine, 0.08 g/L uridine). Strains were then diluted to a final calculated OD of 0.01 in YPAD containing 1% dextrose and added to a 96-well assay plate at a 1:10 ratio, resulting in a final starting OD of 0.001 in each well. Assay plates were prepared by performing serial dilutions of YPAD containing 1% dextrose and 256 μg/mL fluconazole, resulting in final concentrations of 256, 128, 64, 32, 16, 8, 4, 2, 1, or 0.5 μg/mL FLC with cells added, as well as YPAD with 1% dextrose without the addition of FLC.

For MIC₅₀ and SMG assays, 96-well plates were allowed to incubate stationary in a humidified chamber at 30°C for 24 hours, at which time they were resuspended by pipetting and OD₆₀₀ measured using a BioTek Epoch 2 microplate reader. Plates were then allowed to sit stationary for another 24 hours (48 total) at 30°C in a humidified chamber, after which they were resuspended and OD measured. MIC₅₀ was calculated by first subtracting blank values from all wells and then normalizing each strain grown in FLC to the same strain grown in YPAD without FLC and then averaged across replicates. The drug concentration at which average growth was reduced below 50% at 24 hours was designated as the MIC₅₀. SMG was calculated by averaging the normalized growth at 48 hours for all concentrations of FLC greater than the MIC₅₀ value. The triplicate measurements were averaged to produce the final reported SMG value.

For growth curves, 96-well plates containing sterilized water motes surrounding the outer wells were prepared as described above for microbroth dilution assays and grown in a BioTek Epoch 2 at 30°C with shaking for 48 hours, with OD₆₀₀ readings taken every 15 minutes. Data was analyzed in R (version 4.1.3) using the package growthcurver [69] to calculate empirical AUC values. Mean and SE of triplicate OD₆₀₀ values (minus blanks) were calculated and plotted over time using ggplot2 for all growth curve plots.

Evolution experiments

The lineage described in Fig 1 was evolved as previously described in [49]. As detailed there, a single colony isolated from susceptible lab strain SC5314 was suspended in 1 mL sterile YPAD media and grown overnight at 30°C. One fourth of this culture was used to initiate a lineage grown in YPAD + 1 μg/mL FLC in a deep-well 96-well plate which was sealed with Breathe EASIER tape and placed in a humidified chamber at 30°C for 72 hours. Every 72 hours, cells were resuspended and transferred to a new plate with fresh YPAD + 1 μg/mL FLC at a dilution of 1:1000 for a total of 10 transfers. Culture remaining from each passage was frozen at -80°C in 20% glycerol.

In this second experiment, C. albicans cells were evolved in liquid YPD (2% bacto-peptone, 1% yeast extract, 2% dextrose (filter-sterilized)) + 1 μg/mL FLC. The experiment consisted of 4 lineages each initiated from a single colony SC5314 progenitor and passaged at a 1:100 dilution every 24 hours for a total of 12 passages (30°C, 200 rpm), approximating 80 generations. Every 4 days, the passaged cultures were frozen at -80°C in 25% glycerol. The final passage of each lineage was frozen at -80°C in 25% glycerol and selected for further analysis.

Whole genome sequencing, variant detection, and allele frequency analysis

To ensure that the results of whole genome sequencing for evolution passages reflected the entire populations as much as possible, each glycerol stocked passage was thawed and half the entire volume used for a short (~16 hours) growth expansion in 0.5 μg/mL FLC before collection of genomic DNA for sequencing. For all other whole genome sequencing experiments, strains were grown in 3 mL YPAD containing 2% dextrose overnight at 30°C with shaking. Cells were pelleted and resuspended in TENTS buffer (1% SDS, 100 mM NaCl, 100 mM Tris pH 8, 1 mM EDTA, 2% Triton), added to ~250 uL of .5 mm glass beads, and lysed in a BeadRuptor Elite (Omni International, 1 cycle, 4 m/s, 15 s). Genomic DNA was then isolated using a phenol-chloroform extraction. Sequencing libraries were prepared by SeqCenter, LLC, or SeqCoast Genomics using the Illumina DNA Prep and sequenced on an Illumina NovaSeq 6000. Adaptor sequences and low-quality reads were removed using Trimmomatic (v0.39, parameters LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:36 TOPHRED33) [70]. Trimmed reads were mapped to the C. albicans reference genome (A21-s02-m09-r08) from the Candida Genome Database (CGD) (http://www.candidagenome.org/download/sequence/C_albicans_SC5314/Assembly21/). Reads were mapped using BWA-MEM (v0.7.17) with default parameters [71]. Samtools (v1.10) [72] was used to sort, index, and remove PCR duplicates from aligned reads, as well as calculate read depth at each genomic nucleotide position using the ‘samtools depth’ function. Single nucleotide variants were called using GATK (v4.1.2) [73] Mutect2, with SC5314 designated as the ‘normal’ sample to identify new SNVs, or with the mutant designated as the ‘normal’ sample to identify regions of LOH, using default parameters. Variants differing between evolved and progenitor strains were then manually inspected using the Integrated Genome Viewer (IGV, v2.12.3) [74]. Biallelic SNPs for allele frequencies were identified using GATK (v4.1.2) ‘HaplotypeCaller’ [75] followed by ‘GenotypeGVCFs’, both set with -ploidy 2, and then selected using ‘SelectVariants’ with parameters -select-type SNP and–restrict-alleles-to BIALLELIC. Variants were then filtered for quality using QD < 2.0, QUAL < 30.0, SOR > 3.0, FS > 60.0, MQ < 40.0, MQRankSum < -12.5, and ReadPosRankSum < -8.0.

Single nucleotide polymorphisms in the KSR1 gene of 270 whole genome-sequenced C. albicans isolates (S3 Table) were retrieved as fasta files and compared using the MUSCLE Mutliple sequence alignment tool [76]. The impact on the Ksr1 protein sequence was evaluated across isolates using the multiple sequence alignment editor SeaView [77]. All sequences were retrieved from the NCBI Genomes database [78–82].

CNV analysis

Whole genome sequencing was processed as described above, and sequencing depth files generated using ‘samtools depth’ were downloaded and further processed in R (v4.1.3) to identify regions of varying copy number. Depth files were read into R, and rolling means were calculated for 500 bp windows across the entire genome using the RcppRoll R package (v0.3.0) [83]. Means were then divided by the median read depth of the entire genome (excluding mitochondrial DNA) to calculate relative abundance and then multiplied by 2 (due to C. albicans being a diploid organism) to generate an estimated copy number. Estimated copy numbers were then plotted according to genomic position to visualize regions of the genome with changes in copy number. Copy number based on read depth and heterozygosity of the entire genome was also visualized using the online YMAP platform [84]. Raw fastq files were uploaded to YMAP and processed using the parameters “ploidy of experiment: 2, baseline ploidy:2, generate Figs with annotations? Yes, data type: whole genome NGS (short reads), read type: paired-end short reads, reference genome: Candida albicans SC5314 (CGD: A21-s02-m09-r10), haplotype map: Abbey_et_al_2014, chromosome-end bias (forces using GC content bias) checked.”

For mixed populations, such as the entire passage populations of the first evolution experiment, abundance of trisomies or CNVs in the population was estimated using the copy numbers calculated for individual clones isolated from the populations as a reference. The formula A*x + B*(1-x) = C, where B is euploid copy number, A is the CNV copy number estimated from a single colony, and C is the estimated copy number from the entire population, was used to find x, the proportion of the population containing the CNV, assuming no other CNV of a different copy number is present in the same region.

Strain engineering

All primers and sequences used in strain construction and verification are listed in S2 Table. All strains described in this paper (S1 Table) were whole genome sequenced, and the whole genomes sequencing data can be found at the NCBI Sequence Read Archive (BioProject ID PRJNA1071177).

Construction of KSR1 LOH1, KSR1²⁷²*^/*, KSR1B/B, and KSR1A/A strains

KSR1 LOH engineered strains, including KSR1 LOH1, KSR1²⁷²*^/* and KSR1B/B were constructed using homologous recombination. Transformation constructs were generated by amplifying three fragments that were subsequently stitched together into one fragment using strand overlap extension PCR: 1) a 398 bp fragment downstream of the KSR1 locus (relative to the Crick strand) amplified from genomic DNA of the evolved strain containing the KSR1 LOH, 2) a NAT resistance marker optimized for C. albicans from plasmid pAS3019, and 3) the entire KSR1 coding sequence and 464bp immediately upstream (relative to the Crick strand) amplified from genomic DNA of the evolved strain containing the KSR1 locus. Amplification of the KSR1 locus to generate the transformation construct was not allele-specific, and therefore a mixed pool of transformation constructs was generated that included both KSR1B and the chimeric combination of KSR1A and KSR1B that resulted from LOH1. Transformation was performed in SC5314 by lithium acetate and heat shock, and single colonies isolated. Single colonies were screened by PCR for correct upstream and downstream insertion into the genome. Sanger sequencing was then performed for the entire KSR1 locus specifically for the transformed allele by amplifying fragments from within the NAT resistance marker to determine the combination of A and B SNVs present. The genotype of the non-transformed allele was determined by amplification the KSR1 locus, selection of the smaller band (excluding NAT), and restriction digest with restriction enzyme BanII, which will cut the A allele sequence at nucleotide 192 (GAGCCC) but not the B allele sequence (GAACCC). In this way, three strains were selected: 1) one in which the entire A allele had been transformed with the new chimeric allele, resulting in the identical genotype to the evolved strain at the KSR1 locus, 2) one in which the A allele had been partially transformed with the B allele construct, resulting in KSR²⁷²*^/* that was otherwise heterozygous, and 3) one in which the entire A allele had been transformed into a complete B allele at the KSR1 locus resulting in KSR1B/B. Whole genome sequencing was then performed to ensure the correct genotypes at KSR1 and ensure that these strains did not contain additional changes such as aneuploidies or other LOH. The same process was used to generate the KSR1A/A strain, using genomic DNA from the wild-type SC5314 to amplify the KSR1A locus.

Construction of tetO-NCP1 strains

Strains containing an tetO-NCP1 overexpression construct were generated by first amplifying a transformation construct using primers with homology to the native NCP1 locus from a tetO vector pAS3027. The transformation construct, including upstream and downstream homology, included CaTar-TetO-FlpNAT. This construct was transformed into lab strain Sn152, an auxotrophic derivative of SC5314, by lithium acetate and heat shock transformation. Transformants were screened for correct insertion using PCR for both upstream and downstream flanks of the transformed region. Sanger sequencing was used to determine which allele of NCP1 was overexpressed. For strains containing both the tetO-NCP1 and KSR1 LOH1 constructs, NAT was removed from the tetO-NCP1 strains by growth on FBS to activate FLP recombinase, and then transformed with the same protocol described above for KSR1 LOH1.

Construction of KSR1 deletion mutants in SC5314

The KSR1 deletion mutants in the SC5314 background were constructed using the CRISPR system developed by [85]. The intact CRISPR cassette was built with plasmids pADH118 for gRNA expression and plasmid pADH99 for Cas9 components. Both plasmids had the carbenicillin selection marker, and when cut and stitched together they reconstituted the NAT marker. The LEUpOUT strategy was used to remove the CRISPR components and the NAT marker. The C. albicans strain AHY940 (SC5314 LEU2 heterozygous knockout) was used to transform and produce the mutants. The parental strain and plasmids were provided by the Hernday lab.

Construction of strains carrying mNeon-Green-Ksr1p

To generate C. albicans strains with a tagged Ksr1 protein, we amplified: NAT1 from plasmid pJB-T155 (primers 2313 and 2251), mNeonGreen from plasmid pJB-T409 (primers 2315 and 2369), the plasmid backbone containing an E. coli origin of replication, and the Amp^R gene from pJB-T155 (primers 2325 and 2027). The promoter sequence of TDH3 was amplified from the genomic DNA of C. albicans SC5314 (primers 2239 and 2314).

The plasmid expression cassette (NAT1-TDH3-mNeonGreen) was amplified from pJB-T510 by PCR using PCRBIO VeriFi Mix with primers CB-KSR1-F, CB-KSR1-R, which are complementary to 40 bp 5’-to the start codon of the KSR1 ORF, and 40 bp downstream of the start codon. The amplified PCR product was used to transform a fresh C. albicans SC5314 log phase liquid culture and transformants were selected on YPD + nourseothricin (400 μg/ml).

Successful transformants were validated by diagnostic PCR using PhireGreen Hot Start II PCR Master Mix along with primers specific for the expression cassette and KSR1 (#2039, #CB-KSR1-CHK). To determine which allele of the gene was fused to mNeonGreen, the sequence was amplified by PCR (primers #2039, #2342), cleaned using a Qiagen PCR cleaning kit, and Sanger sequenced.

Construction of KSR1^189R/R

The KSR1^189R/R strain was constructed by homologous recombination of a mutated amplicon (generated by oligo-directed mutagenesis) marked with NAT into the progenitor SC5314 background. The mutated amplicon was generated by oligo-directed mutagenesis. Specifically, the second half of the KSR1A allele and the NAT selection marker were amplified from the KSR1A/A strain background using primer 1624 and primer 2042. Primer 2042 contains the nucleotide change KSR1:A566G which codes for amino acid arginine at position 189 in the KSR1B allele. This construct was transformed into SC5314 using heat shock and transformants were selected on nourseothricin plates. Single colony transformants were then screened for reduced filamentation at 37C to select for colonies in which the A allele had been modified. Colonies with reduced filamentation were then screened for correct insertion of the construct by PCR using primers 1660, 1661, 1659, and 1154 (S2 Table) and amplicons Sanger sequenced. Transformants with correct construct sequences were then whole genome sequenced to verify the correct sequences of KSR1A and KSR1B alleles. Strain AMS6412 was selected due to no additional LOH or aneuploidies detected by whole genome sequencing. MIC assays performed on this strain were performed with a different FLC lot than previous MIC assays. The MIC/SMG exactly matched the MIC of the KSR1 LOH1 strain when done in parallel with the same FLC lot. However, the MICs of both strains and controls were lower than the assays done with the previous lot, with an MIC value of 2.

Alphafold protein structure predictions

Ribbon models of structural predictions for KSR1A and KSR1B were generated using Alphafold (version 2.3.2) [62]. Protein coding sequences for KSR1A and KSR1B were downloaded from the Candida Genome Database [86] from the V22 reference genome and submitted for Alphafold predictions using the colab notebook found at: https://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb, using default parameters.

Fluorescence microscopy of tagged KSR1 alleles

Cells were picked from an agar plate into SDC medium and grown overnight by incubating at 30°C with shaking. The following day, cells were diluted 1:50 into fresh SDC. For fluorescence microscopy of the mNeonGreen-tagged protein and staining with an endoplasmic reticulum-specific probe 2 [64] the cells were incubated at 30°C for 2 hours with shaking, the stain was added to a final concentration of 1 μM, and the cells were incubated for an additional hour in the dark. 3–5μL of cell culture was deposited on a glass slide topped with 10 μL of low melt agar and imaged with a Nikon Ti Eclipse microscope using the Nikon Elements AR software, and Nikon Plan Apo x100 objective. Filters used were: 488 nm mNeonGreen excitation and 510 nm for emission; for the ER-specific stain CFP 438 nm excitation and 483 nm emission.

Lipid analysis

Strains AMS2401, AMS5778, AMS5779, AMS5784, AMS5780, and AMS5782 were grown overnight in 10 ml of SDC at 30°C under agitation. Cell suspensions were centrifuged, washed in sterile PBS and counted. Then, 1x10⁸ cells were inoculated in 10 ml of fresh SDC only, or in media containing 1 or 10 μg/ml of fluconazole and grown for additional 48 h at 30°C. Prior to cell lysis, C17-sphingolipid standards were added to the samples [87,88]. Mandala extraction was carried out as described previously [89], followed by Bligh and Dyer Extraction [90]. A third of each sample obtained from the Bligh and Dyer Extraction was reserved for inorganic phosphate (Pi) determination, so the relative sphingolipid signal was normalized by the Pi abundance. The organic phase was transferred to a new tube and submitted to alkaline hydrolysis of phospholipids [91]. Finally, the organic phase was dried and used for mass spectrometry analysis [88]. Sphingolipid abundances, normalized by Pi abundance, were then analyzed using R (version 4.1.3). Any lipid species that were not detected in any sample were removed from analysis. Because the distribution of lipid abundances was exponential across samples, log₁₀ values were calculated for all normalized abundances, and heatmaps generated using the log₁₀ transformed values. Heatmaps were generated using the R package pheatmap (version 1.0.12) [92], and scaled by row (across samples) to result in heatmap color intensities that represent each sample’s z-score calculated from the distribution of all samples in the row. Two samples that clustered away from all other samples with a cluster dendrogram height of >8 were removed as outliers.

RNA sequencing

C. albicans cells were cultured overnight in liquid YPD medium (2% bacto-peptone, 1% yeast extract, 2% dextrose (filter-sterilized)) at 30°C with shaking (200 rpm) and then diluted 1/50 in YPD either without FLC, with 2 μg/ml FLC, or with 64 μg/ml FLC. After 24 h at 30°C with shaking (200 rpm), 1 ml of the cell suspension was utilized to extract total RNA using the Ribopure-Yeast RNA kit (AM1926, Invitrogen), according to the manufacturer’s instructions. A Bioanalyzer (Agilent) was used for the qualitative sample analysis, and only those with RNA quality (RIN) scores of 7 or higher and a 260/280 ratio within the range of 2.13–2.2 were sequenced. Sequencing was performed with the help of the Biomics Platform at Institut Pasteur using pair-end Illumina stranded-mRNA sequencing. RNA sequences reported in this paper have been deposited in the NCBI Sequence Read Archive, https://www.ncbi.nlm.nih.gov/bioproject (BioProject ID PRJNA1063495).

RNA sequencing data analysis

The paired reads from the RNA-seq libraries were trimmed for low-quality reads, and Illumina TruSeq adapters were removed with Cutadapt v1.9.1 [93] with the following parameters:—trim-qualities 30 -e (maximum error rate) 0.1—times 3—overlap 6—minimum-length 30. The cleaning of rRNA sequences was performed with Bowtie2 v2.3.3 [94] with default parameters. The cleaned reads from RNA-seq paired-end libraries were aligned to the C. albicans SC5314 reference genome (Version_A22-s07-m01) with Tophat2 v2.0.14 [95] and the following parameters: minimum intron length 30; minimum intron coverage 30; minimum intron segment 30; maximum intron length 4000; maximum multihits 1; and microexon search. Genes were counted using featureCounts v2.0.0 [96] from the Subreads package (parameters: -t gene -g gene_id -s 2 -p). Analysis of differential expression data was performed in DeSeq2 [97]. The statistical analysis process includes data normalization, graphical exploration of raw and normalized data, tests for differential expression for each feature between the conditions, and raw P-value adjustment. The genes with adjusted P-value less than 0.1 were considered as differentially expressed compared to the non-treated condition (YPD no FLC). Log₂ fold changes for each evolved strain relative to the progenitor strain in no FLC were plotted as heatmaps using pheatmap [92] without any additional scaling, and asterisks were used to denote significant P-values as described above.

Cy5 and R6G flow cytometry

C. albicans cells were grown overnight in liquid YPD medium (1% yeast extract, 2% peptone, and 2% dextrose (with the dextrose filter sterilized)) at 30°C with shaking (200 rpm). Cultures were diluted 1 in 100 in a 3 mL volume of YPD and compounds were added to final concentrations of 1 μg/mL for Rhodamine 6G (Sigma), or 1 μM for FLC-Cy5. Samples incubated with 10 μg/mL of Beauvericin (Sigma) were used as positive controls and those incubated with corresponding volumes of DMSO were used as negative controls. Cells were incubated at 30°C with shaking and harvested after 4 h, washed twice with sterile PBS and diluted to a density of ~10⁶ cells/mL in PBS. Fluorescence was measured in 200 μL aliquots of cell suspension and data were collected from 100,000 cells per sample using a Miltenyi MACSQuant Analyzer 10 Flow Cytometer. Cell populations were gated by SSC/FSC to eliminate small debris particles. Experiments were performed with at least three biological and two technical replicates. Analyses were performed using FlowJo 10.8.

Rhodamine 6G efflux assay

C. albicans cells were patched from glycerol stocks onto YPAD agar plates and grown at 30°C for ~1 day. Patches were used to inoculate 3 mL liquid YPAD media with 2% dextrose (20 g/L bactopeptone, 10 g/L yeast extract, 0.04 g/L adenine, 0.08 g/L uridine), and grown at 30°C overnight with shaking. Cultures were then back-diluted, adding 50 μL of overnight culture into 5 mL of fresh media with 0 μg/mL FLC, 1 μg/mL FLC, or 10 μg/mL FLC added at grown for 3 hours at 30°C with shaking. Cultures were spun down and washed with PBS (pH 7, room temperature), and then resuspended in 2 mL PBS and incubated for 2 hours at 30°C with shaking. Rhodamine 6G (ThermoFisher #419010050) was added to a final concentration of 10 μg/mL and cultures incubated for 1 hour at 30°C with shaking. Cultures were spun down and washed twice with 4°C PBS, and then resuspended in a final volume of 900 μL of PBS. Cultures were then split across 6 wells of a 96-well plate to constitute technical triplicate for a set with and without dextrose (at 150 μL per well). Baseline fluorescence and optical density (OD₆₀₀) measurements were taken using a BioTrex HTX plate reader (excitation 485 nm, emission 525 nm) for 5 minutes at 1 minute intervals. The plate was then removed, and dextrose was added to the ‘with dextrose’ samples to a final concentration of 1%. The plate was then returned to the plate reader and fluorescence and OD measurements were taken every 4 min for a total of 90 min. Three biological replicates were performed.

Fluorescent imaging of lipid droplets

C. albicans cells were cultured in liquid YPD medium (2% bacto-peptone, 1% yeast extract, 2% dextrose (filter-sterilized)) overnight at 30°C with continuous shaking (200 rpm). The number of cells was determined by measuring optical densities (OD₆₀₀) using a Biotek Epoch 2 plate reader. 106 cells/mL were resuspended in YPD in different conditions: no drug, with 128 μg/ml Fluconazole (FLC, PHR1160-1G, Sigma Aldrich), or with 128 μg/ml myriocin (MYO, M1177-5M, Sigma Aldrich). The cell suspensions were incubated for 3 h at 30°C (200 rpm). After incubation, cells were washed twice with sterile PBS and incubated with 1 μg/ml Bodipy 493/503 (D3922, Invitrogen) for 10 minutes at 30°C and 200 rpm. The cell suspension was washed twice with sterile PBS. The cells were then imaged at a 60X magnification with a GFP filter and an exposure of 125 ms on a Zeiss AxioVision Rel. 4.8 microscope. Cellprofiler4.2.6 (https://cellprofiler.org/, available at https://github.com/CellProfiler/CellProfiler) was used to analyze the images. The Speckle Counting pipeline (https://cellprofiler.org/examples) was used to identify smaller objects within larger ones and to establish a relationship between them. Objects with a diameter of 30–100 pixels were considered as yeast cells, while those with a diameter of 1–6 pixels were classified as lipid bodies. The same pipeline was used to quantify several parameters. These include the number of lipid bodies within each cell, the highest pixel intensity within each lipid body (referred to as maximum brightness), and the integrated intensity of lipid bodies. The integrated intensity was quantified by summing the pixel intensities within these structures. All assays were performed with biological triplicates, with at least 200 cells analyzed per replicate.

Fluorescent imaging of vacuoles

C. albicans cells were cultured in liquid YPD medium (2% bacto-peptone, 1% yeast extract, 2% dextrose (filter-sterilized)) overnight at 30°C with shaking (200 rpm) and resuspended in YPD or YPD + FLC (128 μg/ml) at a concentration of 106 cells/mL. To stain the vacuoles, cells were incubated with 8 μM of SynaptoRed C2 (S6689, Sigma) for 3 h at 30°C with shaking. The cell suspension was then washed with 1 mL of fresh YPD and transferred to 4 mL of YPD before incubating for 90 min at 30°C (200 rpm). The cells were washed with sterile PBS before imaging at a 60X magnification with an mCherry filter and an exposure of 100 ms on a Zeiss AxioVision Rel. 4.8 microscope. Cellprofiler4.2.6 (https://cellprofiler.org/, available at https://github.com/CellProfiler/CellProfiler) was used to analyze the images. The Speckle Counting pipeline (https://cellprofiler.org/examples) was used to identify objects with diameters of 10–35 pixels, these objects were considered as vacuoles. The sum of the pixel intensities within these vacuoles, represented by the integrated intensity, and the area of these vacuoles were quantified using Speckle Counting. All assays were performed with biological triplicates, with at least 200 cells analyzed for each replicate.

RT-qPCR

C. albicans cells were patched onto solid YPAD agar from frozen glycerol stocks and grown at 30°C for ~2 days. Patches were used to inoculate 3 mL YPAD media (2% dextrose) and grown overnight at 30°C with shaking. 350 μL of overnight cultures were back diluted into 50mL fresh YPAD media with 0 μg/mL FLC or 1 μg/mL FLC added (and/or doxycycline supplemented with iron for tetO-NCP1 strains) and grown at 30°C with shaking until cultures reached an OD between 0.4–0.5. Each culture was pelleted, supernatant removed, and cell pellets flash frozen in liquid nitrogen, and then stored at -80°C until RNA could be harvested. RNA was extracted using the Qiagen RNeasy Mini Kit (CAT# 74106), using the mechanical disruption method with an Omni International Bead Ruptor Elite. 1 ug of RNA from each sample was DNase treated after elution, via incubation with DNase at 37°C for 30 min followed by the addition of EDTA to a final concentration of 5mM and heat inactivation at 75°C for 5 min. cDNA was prepared using SuperScript II RT (ThermoFisher, Cat# 18064014) using an Oligo(dT) primer according to the manufacturer’s recommendations. cDNA was used as template for qPCR reactions performed in technical triplicate with primers specific to either TEF1, ACT1, NCP1, or CDR1 for each sample as well as noRT and no template controls, using PowerUp SYBR Green Mastermix (Fisher Scientific, CAT# A25742) run on a BioRad CFX-96. Delta-delta Ct values for CDR1 were calculated, with propagation of error, relative to either TEF1 or ACT1 and then relative to the progenitor strain in 0 μg/mL FLC. Delta-delta Ct values for NCP1 were calculated with propagation of error relative to ACT1 and then relative to the progenitor strain in 0 μg/mL FLC. Three complete biological replicates were performed.

Spot plates

C. albicans cells were patched onto solid YPAD agar plates and grown at 30°C for ~2 days. Patches were used to inoculate 3 mL of liquid YPAD media (2% dextrose) and grown overnight at 30°C with shaking. Cultures were diluted to ~10⁷ cells/mL in PBS and serial dilutions performed to generate suspensions at 10⁶, 10⁵, 10⁴, and 10³ cells/mL. 5 μL of each suspension was spotted onto YPAD agar plates with 0% or 0.02% sodium dodecyl sulfate. Cells were incubated at 30°C and pictures taken at 24 and 48 hours.

Filamentation assays

C. albicans cells were cultured in liquid YPD medium (2% bacto-peptone, 1% yeast extract, 2% dextrose (filter-sterilized)) overnight at 30°C with shaking (200 rpm). Yeast cells were diluted 1/50 in 4 ml of YPD and seeded in 6-well plates. After 2 h at 37°C with shaking (200 rpm), the cell suspension was then washed twice with sterile PBS and imaged at a 60X magnification in brightfield on a Zeiss AxioVision Rel. 4.8 microscope. All assays were performed with three biological replicates, with at least 200 cells analyzed for each replicate. Representative images are shown in the Figs.

Galleria mellonella infections

G. mellonella larvae were sourced from La ferme aux Coleos (Cherbourg-en-Cotentin, France), maintained at room temperature, and were used within one week of delivery. Larvae without any signs of melanization and with an average weight of 0.4 g were selected for each experiment in groups of 12. C. albicans cells were cultured in liquid YPD medium (2% bacto-peptone, 1% yeast extract, 2% dextrose, filter-sterilized). Following an overnight incubation at 30°C with shaking (200 rpm), cells were washed twice with sterile PBS. The cell concentrations were then determined using a Luna FX7 Automatic Cell Counter (Logos Biosystems) and adjusted to 3 × 10⁷ cells/mL in sterile PBS. For the infection, each larva was injected with 3 × 10⁵ cells/mL through the last left pro-leg using a 10 μL glass syringe and a 26S gauge needle (Hamilton, 80300). A subsequent injection with either fluconazole (0.2 μg/larva) or a sterile vehicle (PBS at matching injection volumes) was administered via the last right pro-leg, 2 h post-infection. The inoculum size was verified by plating fungal cells on YPD and counting the resulting colony forming units (CFUs). Both infected and control groups of larvae were maintained at 37°C for 14 days, with daily monitoring of survival. Larvae were deemed dead if no movement was observed upon contact. The experiments were conducted with three biological replicates. Control groups of larvae included those treated with PBS or FLC without fungal infection.

Amphotericin B e-strip sensitivity tests

Cells were grown on SAB-dex agar plates at 35C for 24 hours from glycerol stocks. Cells were then resuspended in PBS and normalized to an OD of 0.1–0.15. 100μL of the normalized cell suspension was then plated on RPMI agar plates using glass beads and an Amphotericin B e-strip (Liofilchem Ref 92153, lot 031224125) placed in the center of the plate. Plates were photographed and read after 24 hours of growth at 35C in a humidified chamber.

Sterol analysis

Cells were grown overnight in 50mL 2% dextrose YPAD media at 30C with shaking. Cells were collected by centrifugation and washed once with PBS and resuspended in 25mL PBS. 500μL of the cell suspension was then diluted into 50mL fresh CSM media (1.7 g/L YNB w/out ammonium sulfate, 5 g/L ammonium sulfate, 20 g/L glucose, 2 g/L SC AA mix) with 0, 1, or 10 μg/mL FLC added and allowed to grow for 5 hours at 30C with shaking. After 5 hours the OD600 was used to normalize all samples to a total cell number of ~5x10^8 cells. These cells were harvested by centrifugation, washed once in PBS, and transferred to 1.5mL Eppendorf tubes, pelleted by centrifugation, supernatant discarded, and cell pellets frozen at -80C. Cells were then shipped from Minneapolis to Stony Brook on dry ice for processing. Once there, lipid extraction and GC-MS was conducted as previously described [98]. Pellets with 5x10⁸ cells were used for lipid extraction. The dried total samples were resuspended in 100 μL chloroform added to 100 μL of BSTFA reagent (Thermo Scientific) and incubated at 70°C for 1 hour prior to GC-MS (Agilent 7890B GC–MS, Agilent 5977A MSD) analysis [88]. The retention time and mass spectral patterns of a sterol standard were used as references for lipid analysis. Sterol standards used in this study include ergosterol (Smolecule, catalog #S527372), lanosterol (Smolecule, catalog #S532452), obtusifoliol (Smolecule, catalog #S563624), zymosterol (Smolecule, catalog #S580329), 4,4-dimethyl zymosterol (Avanti catalog #700073), eburicol (Smolecule, catalog #S633611), episterol (Smolecule, catalog #S628882), and gramisterol (24-methylenelophenol or 4-methyl episterol) (Smolecule, catalog #S626191). Cholesterol (Avanti catalog # 700100) was added as an internal standard for these analyses prior to lipid extraction. The relative amount of unknown sterols was estimated based on the relative percentage of the sterol to ergosterol peak areas in each sample. The mass spectrum of the sterol likely to be the toxic dienol was compared to the authentic standards and spectra profiles from the National Institute of Standards and Technology search database for 4-methyl episterol.

Mouse survival study

Male CD-1 mice were inoculated via the tail vein with 2 x 10⁵ cells of C. albicans, using 5 mice per strain. The mice were monitored twice daily for survival by an observer who was blinded to the identity of the infecting strain. Mice that showed signs of substantial distress were humanely euthanized.