(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Evolutionary dynamics in gut-colonizing Candida glabrata during caspofungin therapy: Emergence of clinically important mutations in sphingolipid biosynthesis [1]

['Yasmine Hassoun', 'Hackensack Meridian Health Center For Discovery', 'Innovation', 'Nutley', 'New Jersey', 'United States Of America', 'Ariel A. Aptekmann', 'Mikhail V. Keniya', 'Rosa Y. Gomez', 'Nicole Alayo']

Date: 2024-09

Invasive fungal infections are associated with high mortality, which is exacerbated by the limited antifungal drug armamentarium and increasing antifungal drug resistance. Echinocandins are a frontline antifungal drug class targeting β-glucan synthase (GS), a fungal cell wall biosynthetic enzyme. Echinocandin resistance is generally low but increasing in species like Candida glabrata, an opportunistic yeast pathogen colonizing human mucosal surfaces. Mutations in GS-encoding genes (FKS1 and FKS2 in C. glabrata) are strongly associated with clinical echinocandin failure, but epidemiological studies show that other, as yet unidentified factors also influence echinocandin susceptibility. Furthermore, although the gut is known to be an important reservoir for emergence of drug-resistant strains, the evolution of resistance is not well understood. Here, we studied the evolutionary dynamics of C. glabrata colonizing the gut of immunocompetent mice during treatment with caspofungin, a widely-used echinocandin. Whole genome and amplicon sequencing revealed rapid genetic diversification of this C. glabrata population during treatment and the emergence of both drug target (FKS2) and non-drug target mutations, the latter predominantly in the FEN1 gene encoding a fatty acid elongase functioning in sphingolipid biosynthesis. The fen1 mutants displayed high fitness in the gut specifically during caspofungin treatment and contained high levels of phytosphingosine, whereas genetic depletion of phytosphingosine by deletion of YPC1 gene hypersensitized the wild type strain to caspofungin and was epistatic to fen1Δ. Furthermore, high resolution imaging and mass spectrometry showed that reduced caspofungin susceptibility in fen1Δ cells was associated with reduced caspofungin binding to the plasma membrane. Finally, we identified several different fen1 mutations in clinical C. glabrata isolates, which phenocopied the fen1Δ mutant, causing reduced caspofungin susceptibility. These studies reveal new genetic and molecular determinants of clinical caspofungin susceptibility and illuminate the dynamic evolution of drug target and non-drug target mutations reducing echinocandin efficacy in patients colonized with C. glabrata.

Invasive fungal infections cause high mortality due to our limited antifungal drug armamentarium and increasing antifungal drug resistance. Echinocandins are a frontline antifungal class with increasing resistance in Candida glabrata, an opportunistic pathogen colonizing host mucosal surfaces. It is known that echinocandin-resistant C. glabrata isolates emerge from drug-sensitive gut-colonizing strains, but there are large knowledge gaps regarding how this resistance develops and whether it involves mutations in genes other than those encoding the echinocandin drug target. We studied the evolutionary dynamics in C. glabrata colonizing the mouse gut during treatment with caspofungin, a widely used echinocandin. In addition to well-described drug target mutations, we identified mutations in sphingolipid biosynthesis gene FEN1 rapidly and frequently emerging in the gut-colonizing fungus during caspofungin therapy. We also identified multiple loss-of-function fen1 mutations in clinical C. glabrata isolates and showed that they contribute to reduced caspofungin sensitivity of these strains. Together, this work illuminates the rich evolutionary dynamics of gut-colonizing fungi during antifungal therapy and identifies a new genetic determinant of reduced clinical caspofungin susceptibility.

In this study, we examined the evolution of resistance to the echinocandin caspofungin in C. glabrata colonizing the mouse gut. We found that both fks and non-fks mutations arose in GI-colonizing C. glabrata during treatment with a humanized dose of caspofungin. The majority of non-fks mutations were in the FEN1 gene, which encodes a fatty acid synthase and functions in sphingolipid biosynthesis. Further analysis demonstrated that fen1 mutants have a fitness advantage in the gut specifically during caspofungin treatment, that these mutants accumulate phytosphingosine (PHS, a sphingolipid biosynthesis intermediate), and that genetic depletion of PHS by deletion of alkaline ceramidase YPC1 sensitizes both the wild type strain and the fen1 mutant to caspofungin. Interestingly, in-depth analysis of the emergence and abundance of both drug target (fks) and non-drug target (fen1) mutations during caspofungin treatment by amplicon sequencing revealed a rapid emergence and diversification of caspofungin-adapted strains in the GI tract. Importantly, we also identified three different point mutations in FEN1 in C. glabrata clinical isolates and showed that all of these clinical fen1 mutations cause loss of protein function and reduced susceptibility to caspofungin. In summary, our study uncovered mutations in FEN1 as a novel genetic mechanism reducing the susceptibility of C. glabrata to the widely used echinocandin caspofungin in the clinical setting and showed that this effect is due to an increase in the membrane level of PHS. Beyond its clinical implications, our work also sheds light on the complex evolutionary dynamics of C. glabrata within the host GI reservoir, comprising both drug target and non-target mutations, which collectively undermine the effectiveness of antifungal treatment in patients colonized with C. glabrata.

A number of studies have attempted to elucidate the genetic underpinnings of echinocandin resistance in C. glabrata by comparing whole genome sequences of resistant and susceptible clinical isolates [ 24 , 30 – 34 ]. This approach is complicated by the high genetic diversity of C. glabrata, where thousands or even tens of thousands of SNPs and indels can separate different clinical isolates [ 35 ]. Thus, genome comparisons typically identify many putative genetic changes regulating echinocandin sensitivity, but validating the role of each individual SNP is extremely labor-intensive and usually not carried out, nor are the mechanisms by which these SNPs may alter echinocandin sensitivity elucidated. In another approach to identify the missing genetic basis for echinocandin resistance, in vitro evolution has been used to show that C. glabrata passaged in rich laboratory medium containing increasing concentrations of anidulafungin acquires mutations in FKS1 and FKS2 as well as mutations in ergosterol biosynthesis gene ERG3, implicating the latter in echinocandin resistance via as yet unknown mechanisms [ 36 ]. However, it is not yet clear how relevant this narrow mutational signature of resistance is to the mutations driving the evolution of echinocandin resistance in the context of the host. Importantly, the gastrointestinal (GI) tract and the abdominal cavity were identified as key host reservoirs in which C. glabrata drug resistance emerges [ 37 – 39 ] and in vitro growth in rich medium may not recapitulate the evolutionary dynamics of C. glabrata existing within these host niches.

Echinocandins, which are fungicidal in Candida, target the biosynthesis of the fungal cell wall by inhibiting the enzyme β-glucan synthase [ 25 ]. The best understood echinocandin resistance mechanism is via mutations in genes encoding β-glucan synthase (FKS1 and FKS2 in C. glabrata) that alter the structure of the enzyme, rendering it less susceptible to inhibition [ 14 , 26 , 27 ]. While fks mutations have been firmly linked to C. glabrata echinocandin resistance leading to clinical failure, there is also evidence for a contribution of additional, as yet unidentified genetic mechanisms. For instance, in many epidemiological studies a significant proportion (up to 40%) of resistant C. glabrata isolates do not contain fks mutations [ 16 – 18 , 28 ]. Furthermore, although the mechanism of action of different echinocandins (caspofungin, micafungin, and anidulafungin) is thought to be identical, it is not uncommon for C. glabrata strains to show resistance to one or two but not all three of these drugs [ 15 – 18 , 29 ]. Thus, it is clear that there exist additional, as yet unknown, important determinants of susceptibility of C. glabrata to all echinocandins as well as to individual drugs of this class.

Invasive fungal infections cause a significant health burden worldwide, with associated mortality equaling or surpassing that of malaria and TB [ 1 – 4 ]. Invasive Candida infections are associated with a mortality of >40% even with antifungal treatment [ 3 , 5 , 6 ]. The epidemiology of invasive candidiasis has been changing globally, with formerly predominant C. albicans decreasing in prevalence, while non-albicans Candida species, such as C. glabrata (recently reclassified as Nakaseomyces glabratus), C. parapsilosis, and C. tropicalis now account for ever greater fractions of invasive Candida infections [ 5 , 7 , 8 ]. Alarmingly, the non-albicans Candida species are characterized by reduced susceptibility to the azole antifungal drug class, which has been a global staple of antifungal therapy for several decades [ 9 ]. It has been proposed that the extensive use of azoles, both in medicine and in agriculture, has driven this epidemiological shift [ 10 ]. Due to this changing epidemiology, the other frontline antifungal drug class, the echinocandins, is becoming the treatment of choice for invasive candidiasis. Echinocandins have an excellent safety profile and are highly efficacious against Candida infections [ 11 , 12 ]. However, as echinocandins have been used clinically for over 20 years, resistance to this antifungal drug is also rising, most prominently in C. glabrata [ 13 – 18 ], where many cases of rapid emergence of echinocandin resistance during therapy have been documented [ 19 – 24 ].

Although the fen1-G143V mutation was previously reported, it was not known whether it affected Fen1 function, and the same was true for the newly discovered fen1-G52A mutation. Because strains WM_18.26 and CAS14-5989 were not readily available, we introduced the G143V and G52A mutations, separately, into FEN1 carried on a low copy plasmid and transformed these plasmids into the fen1Δ strain. Unlike the plasmid carrying wild type FEN1, which fully complemented the loss of Fen1 by restoring wild type CSF sensitivity, the plasmids carrying fen1-G143V or fen1-G52A behaved like the empty vector, strongly suggesting that the fen1-G143V mutation caused a loss of Fen1 function ( Fig 6B ). To determine whether the reduced CSF sensitivity of strain DPL239 was due to the fen1-Y298* mutation, we transformed it either with the FEN1 plasmid or an empty vector. Whereas the empty vector had no effect, the FEN1 plasmid sensitized the strain to CSF ( Fig 6B ), strongly suggesting that the fen1-Y298* mutation contributes to the reduced CSF susceptibility of this clinical C. glabrata isolate. The same result was obtained with gut-evolved strain GI-240 ( Fig 6B ), confirming that it contained a loss-of-function fen1 mutation ( Fig 1A ). Together, these experiments showed that loss-of-function fen1 mutations occur in clinical C. glabrata strains and contribute to the strains’ reduced susceptibility to CSF.

A. Analysis of publicly available C. glabrata genomes and Sanger sequencing of FEN1 in the Perlin lab strain collection identified three unique (non-ST-associated) fen1 point mutations in clinical strains. Two of the strains also contained mutations in either FKS1 or FKS2. The MIC values for CSF and micafungin (MCF) for WM_18.26 were obtained from [ 30 ], for DPL239 from our measurements, and for CAS14-5989 from CDC records. B. All three identified clinical fen1 mutations phenocopy fen1Δ and reduce C. glabrata CSF sensitivity. Strains containing wild type FEN1 do not grow in the presence of CSF, whereas strains lacking either functional chromosomal or plasmid FEN1 grow in the presence of the drug. When introduced into C. glabrata on a low-copy plasmid, neither fen1 G143V nor fen1 G52A were able to complement fen1Δ CSF-resistant phenotype. Conversely, the CSF-resistant phenotype of clinical and gut-evolved strains carrying chromosomal fen1 mutations was complemented by WT FEN1 introduced on a low-copy plasmid.

To investigate whether fen1 mutations also occur in C. glabrata isolates infecting humans, we used two approaches: mining publicly available genomes and sequencing FEN1 in strains from our lab collection. We used two sources of clinical isolate genome data: those in Biswas et al. (52 strains) [ 30 ] and those collected by the Centers for Disease Control (CDC) as part of the Emerging Infections Program (EIP) for candidemia surveillance and made publicly available on NCBI [ 67 ]. Because strains of different sequence types (ST) are expected to contain ST-specific polymorphisms in FEN1, we also determined the ST of each strain, either from the whole genome sequence or by Sanger sequencing using the standard C. glabrata multi-locus sequence typing scheme [ 54 ]. These analyses identified three clinical C. glabrata isolates carrying unique mutations in FEN1 that were not ST-associated polymorphisms: G143V in strain WM_18.26 [ 30 ], Y298* in strain DPL239 (Perlin lab collection), and G52A in strain CAS14-5989 (CDC collection) ( Fig 6A and S3 and S4 Tables). WM_18.26 also had the fks2-S663P mutation and a high reported CSF MIC (16 μg/ml), while CAS14-5989 also contained the fks1-S595T mutation but a low CSF MIC (0.03 μg/ml, according to CDC records) ( Fig 6A ). DPL239 with fen1-Y298* contained wild type FKS1 and FKS2 sequences but an elevated CSF MIC (1 μg/ml). The combination of fen1-G143V with fks2-S663P in the genome of a differently named clinical isolate (CMRL4) has previously been reported [ 34 ]. Because both WM_18.26 and CMRL4 were originally reported by the same group within one year of each other [ 30 , 55 ], it is likely that they are either the same strain or two closely related strains. Interestingly, the CMRL4 strain was obtained from a patient who had had 30 days of CSF treatment [ 55 ], consistent with emergence of the fen1-G143V mutation during CSF therapy. No history of antifungal treatment was available for strains DPL239 and CAS14-5989, but because the latter also contained an fks1 hot-spot mutation, it may be hypothesized that the fen1 and fks1 mutations may have arisen under a similar environmental pressure.

To gain an even deeper insight into the timing and spread of FKS and FEN1 mutations in C. glabrata colonizing the gut during CSF treatment, we used amplicon sequencing ( Fig 5B ). Four mice were colonized with C. glabrata and treated with CSF as in our previous studies [ 37 , 40 ] ( S2A Fig ). Feces were collected on the indicated days and plated on YPD agar plates supplemented with antibiotics, as before, but instead of picking individual yeast colonies, ~1000 colonies were pooled for DNA isolation, followed by amplicon PCR and next generation sequencing. The resulting reads were mapped to the amplicon sequences to identify SNPs appearing in the gut-colonizing C. glabrata ( Fig 5B ). First, we observed that the original inoculum contained several SNPs in FKS1 HS1 present at low frequencies ( S2B Fig ), suggesting that these mutations arose in the C. glabrata culture used for mouse oral gavage. These mutations tended to persist in the mouse gut at similar frequencies throughout the two-week CSF treatment, indicating that they did not affect fitness under these conditions ( S2B Fig ). In contrast, mutations in FEN1 and FKS2 appeared and rapidly increased in abundance in mice 1, 3 and 4 ( Fig 5B ), indicating that they increased fitness. The earliest fen1 mutation (W39*), in mouse 3, was detected on day 7, whereas the earliest fks2 mutation (R1378H), in mouse 1, was detected on day 9 (and in this mouse no fen1 mutation was ever detected). Interestingly, although both of these mutations arose relatively early during the treatment, neither of them swept the population but remained at a <50% abundance ( Fig 5B ), suggesting that they co-existed with other fitness-promoting mutations that had arisen in other cells. Furthermore, the fen1-W39* mutation in mouse 3 started decreasing in abundance by day 13 and was undetected on day 15, at which point the fks2-F659I mutation, which is associated with clinical echinocandin resistance [ 53 ], had appeared in the same population ( Fig 5B ). Finally, although a few FKS1 HS mutations had appeared in mice 3 and 4, none of them significantly increased in abundance or were detected on multiple days, indicating that they had marginal effects on fitness during CSF treatment. Thus, this experiment confirmed that GI-colonizing C. glabrata was highly genetically dynamic during CSF treatment, with caspofungin-adapted mutations promoting fitness arising, increasing in relative abundance, and co-existing with other such mutations.

A. Sanger sequencing of FKS1 and FKS2 hot spots and FEN1 ORF in individual colonies (two per mouse per day) derived from gut-colonizing C. glabrata reveals the wide variety of fen1 mutations, some detected as early as day 7 (i.e., 4 days after initiation of CSF treatment). No FKS1 hot spot mutations were found. B. Amplicon sequencing of FKS1 and FKS2 hot spots and FEN1 ORF revealed the time of appearance and relative abundance of fks1, fks2, and fen1 mutant alleles in gut-colonizing C. glabrata during CSF treatment. Fraction of reads containing the indicated allele relative to all reads containing that position was used as a proxy for the relative abundance of the mutant allele in the gut-colonizing C. glabrata population. Parts of Panels A and B of this figure were made using Biorender.com (lab license).

To check whether the results obtained for the strains chosen for whole genome sequencing ( Fig 1A ) were representative of general C. glabrata evolution in the gut (i.e., that fen1 mutations arose frequently in this population during caspofungin treatment), we used yeast samples obtained and stocked during previous gut colonization experiments ( Fig 5A ). Because this analysis was done post hoc, it was done on a limited number of colonies: on those occasions, two colonies per mouse per day had been cultured, FKS1 and FKS2 hot spots sequenced, and frozen stocks made. We now sequenced FEN1 as well ( Fig 5A ). This analysis confirmed that a variety of fen1 mutations had arisen in the different mice, including two different fen1 mutations co-existing in mouse 4 on day 7. Interestingly, in mice 1, 4, and 5 where both fen1 and fks2 mutations were observed, the fen1 mutations had appeared earlier than fks2 mutations ( Fig 5A ; no fks1 mutations had been detected). In mice 2, 3 and 6 only fen1 mutations were detected, and these also appeared early–on days 7 and 9, i.e., 4 and 6 days after the initiation of CSF treatment. These results suggested that the gut-colonizing population of C. glabrata during CSF treatment is highly dynamic and diverse, rapidly acquiring both drug target and non-drug target mutations that improve fitness in the presence of CSF.

A . Lipidomic analysis showed that phytosphingosine (PHS, red arrowhead) levels are increased in gut-evolved CSF-adapted C. glabrata mutants. The indicated strains were either left untreated (No CSF) or exposed to CSF (0.5 μg/ml) for one hour. The samples were analyzed at the Lipidomics Shared Resource at the Medical University of South Carolina. All samples were normalized against inorganic phosphate and the untreated WT control. B. Deletion of alkaline ceramidase YPC1, which breaks down phytoceramide to produce PHS [ 51 , 52 ], sensitized wild type and fen1Δ C. glabrata to CSF. This sensitization was evident when C. glabrata was grown on solid CSF-containing YPD medium (0.125 μg/ml and 0.25 μg/ml CSF) but not in liquid RPMI, where ypc1Δ did not alter the CSF MIC. Both growth on solid medium and the MIC measurement showed that ypc1Δ was epistatic to fen1Δ, sensitizing it to CSF. MCF = micafungin; ANF = anidulafungin C. Deletion of YPC1 also restored CSF binding to the fungal cell surface in the fen1Δ strain. BF = bright-field; AF = Alexa Fluor.

Previous studies have shown that in C. albicans some sphingolipid biosynthesis mutants accumulated phytosphingosine (PHS), a sphingolipid biosynthesis intermediate [ 44 ], and that increased PHS can render GS less susceptible to inhibition by echinocandins [ 49 , 50 ] . Thus, we performed a sphingolipid-focused lipidomic analysis of two fen1 mutants (fen1Δ and GI-299) and one non-fen1 gut-evolved strain (GI-113) both in the absence and presence of 1 μg/ml CSF. The results showed that all three gut evolved mutants (both fen1 and non-fen1) had elevated levels of PHS, which were increased further during CSF exposure ( Fig 4A and S2 Table ). To further probe the importance of PHS accumulation in regulating CSF susceptibility, we deleted YPC1, the gene encoding alkaline ceramidase, which in S. cerevisiae catalyzes the conversion of phytoceramide to PHS [ 51 , 52 ] ( Fig 1B ). Interestingly, we observed that ypc1Δ had the opposite effect of fen1Δ, hypersensitizing C. glabrata to CSF, which was evident on agar plates containing CSF ( Fig 4B ). Furthermore, ypc1Δ was epistatic to fen1Δ, as the double mutant was as sensitive as the ypc1Δ alone ( Fig 4B ). Consistent with these results, ypc1Δ also restored CSF-AF647 association with the fungal cells in the fen1Δ mutant ( Fig 4C ). We also attempted to delete YPC1 in gut-evolved strain GI-113, but despite multiple attempts, no deletants were recovered, suggesting a possible synthetic lethality with one of the mutations in that strain ( S1 Table ). Together, these results show that accumulation of PHS is associated with reduced CSF sensitivity and reduced CSF binding to the plasma membrane, whereas genetic depletion of PHS, conversely, sensitizes C. glabrata to CSF and enhances CSF association with the plasma membrane.

Unfortunately, the addition of AF647 significantly reduced the antifungal activity of CSF-AF647 relative to CSF: the MIC of CSF-AF647 was ~0.85% of that of CSF in both WT and fen1Δ strains ( S1B Fig ). Furthermore, our mass spectrometry (MS) analysis of the CSF-AF647 preparation showed that it contained 0.6% unlabeled CSF, which may have accounted for some of the observed antifungal activity of CSF-AF647. Thus, the reduced association of CSF with plasma membrane in the fen1Δ mutant needed to be verified using a different approach not reliant on the CSF-AF647 conjugate. To this end, we used MS to directly measure the binding of unlabeled CSF to cellular membranes in WT and fen1Δ strains. C. glabrata cells were incubated with 0.25 μg/ml CSF for 20 minutes, followed by immediate cell harvesting and membrane extraction. 10% the membrane preparation was saved for protein quantification, and the rest was analyzed by LC-MS/MS. The detected CSF levels were internally normalized to protein levels in the same sample. Consistent with the imaging data ( Fig 3B ), we observed a 50% reduction in CSF association with the membranes derived from the fen1Δ mutant relative to the membranes derived from the WT strain ( Fig 3C ). On the other hand, the membrane protein band patterns of the WT and fen1Δ strains were very similar ( Fig 3D ), consistent with the conclusion that the reduced CSF association in the fen1Δ mutant was likely due to its altered membrane lipid composition. In summary, both imaging and MS-based analyses have shown that the interaction of CSF with the plasma membrane is significantly reduced in the fen1Δ mutant, potentially explaining the reduced sensitivity of this mutant to CSF.

A. Alexa Fluor 647 was conjugated to the indicated primary amine group of CSF (red circle). The synthesis was carried out at WuXi Biologics (China). The product was confirmed by nuclear magnetic resonance and high-resolution mass spectrometry and determined to be 98.67% pure. CSF’s fatty acid-like tail is indicated by the dashed blue box. B. High-resolution imaging showed that CSF-AF647 strongly associated with the fungal cell surface in WT cells, but that this association was greatly reduced in the fen1Δ mutant. C. CSF interaction with cellular membranes was significantly reduced in the fen1Δ mutant relative to the WT strain. Cells were incubated with CSF, disrupted, and crude membranes were extracted and analyzed by LC-MS/MS. The detected CSF levels were internally normalized to each sample’s protein levels. Each data point corresponds to a single biological replicate. D. Protein composition of the crude membrane preparations (Coomassie-stained 8% TRIS-Glycine SDS PAGE) was unaltered by fen1Δ. BF = bright-field; AF = Alexa Fluor. *** p<0.001 (two-tailed unpaired T-test).

Sphingolipid biosynthesis has been previously implicated in regulating fungal susceptibility in vitro to CSF but not to other echinocandins [ 43 , 44 ]. However, the molecular basis for the reduced CSF susceptibility of sphingolipid mutants, such as fen1Δ, is not known. It has been previously proposed that the fatty acid-like tail of CSF ( Fig 3A ), which is not present in other echinocandins, may modulate its interaction with the plasma membrane and membrane-bound glucan synthase [ 43 ]. To test this hypothesis, we first analyzed the CSF-plasma membrane interaction using fluorescent CSF. This compound (custom synthesized and purified at WuXi Biologicals, China) was obtained by reacting a free amine group on CSF with Alexa Fluor 647 (AF647) modified with an amine-reactive NHS ester ( Fig 3A ). The WT and fen1Δ strains were incubated with CSF-AF647 and analyzed by high resolution imaging (ONi Nanoimager). The imaging showed that CSF-AF647 showed a strong association with the cellular periphery in the WT cells ( Fig 3B ), consistent with its association with the plasma membrane. Importantly, this association was fully dependent on the CSF moiety of the molecule, as AF647 alone (in which the amino-reactive ester had been quenched with TRIS) did not bind to C. glabrata cells ( S1C Fig ). Interestingly, the binding of CSF-AF647 to the fungal cell surface was significantly reduced in the fen1Δ cells, and this reduction observed when FEN1 was deleted in strains CBS138 ( Fig 3B ) and ATCC90030 ( S1A Fig ), showing that it was independent of C. glabrata strain background. Thus, these observations suggested that a key difference between WT and fen1Δ cells is a decreased association between CSF and the plasma membrane.

We also analyzed the effect of fen1 mutation on C. glabrata CSF susceptibility in the context of a bloodstream infection. Mice were immunosuppressed and infected via the retro-orbital route with either wild type, fen1Δ, or fks2-S663P mutant strains, the latter of which is a prominent cause of clinical echinocandin resistance and served as a control. Kidney fungal burdens were assessed as a measure of CSF efficacy as described before [ 48 ]. We used a range of CSF doses (1.25, 2.5, and 5 mg/kg), the highest of which corresponds to the humanized dose. Consistent with the documented resistance of the fks2-S663P mutant [ 26 ], we found that its burden in the kidney was not reduced as much as those of the wild type strain at all three concentrations, with the difference being statistically significant at 5 mg/kg ( Fig 2C ). In contrast, there was no difference in kidney burdens between the fen1Δ and the wild type strain at any of the caspofungin concentrations ( Fig 2C ), showing that in the context of bloodstream infection, unlike during gut colonization, the fen1Δ mutant remained susceptible to CSF.

A. Gut-colonizing fen1 mutants survive during CSF therapy better than the isogenic wild type strain. The mice were colonized with the indicated C. glabrata strains and treated daily with CSF (20 mg/kg) starting at day 3 post-colonization. Fecal samples were obtained every other day and plated to obtain GI fungal CFU counts. Multivariate Analysis of Variance (MANOVA) was used to assess the relationship between treatments over time on gut microbiota with a significant threshold set at 0.05. Tukey’s Post Hoc analysis was performed to look for pairwise significance. “Fungal burden depletion” refers to C. glabrata gut levels decreasing to undetectable levels during CSF treatment and not rebounding during the timeframe of the experiment. B. In direct competition experiments, gut-colonizing fen1 mutants outcompete the wild type strain specifically during CSF treatment. The mice were colonized with an equal mix of two strains: wild type and an isogenic fen1 mutant (fen1Δ or fen1-W145*). Starting at day 1 post-colonization, the mice were treated daily with CSF (20 mg/kg). Fecal samples were plated every other day, and the resulting colonies were analyzed for their FEN1 status (wild type or mutant). To identify fen1Δ colonies, the plates were replica-plated on hygromycin-containing medium. To identify fen1-W145* colonies, a specific molecular beacon was designed and used in colony PCR (see Materials and Methods ). C. In a model of systemic infection, the fen1Δ mutant was as sensitive to CSF treatment as the wild type strain. Neutropenic mice were infected via the retro-orbital route with the indicated strains of C. glabrata, and treatment with CSF (1.25, 2.5 or 5 mg/kg) or vehicle (PBS) was initiated at day 1 post-infection. The kidneys were harvested on day 3, homogenized and plated on YPD for CFU counts. Each data point corresponds to a single mouse. *** p < 0.001 (ordinary one-way ANOVA).

To directly analyze the role of FEN1 during host colonization and infection, we deleted the FEN1 ORF and colonized mice with fen1Δ and fen1-W145* (GI-240, Fig 1A ). We found that daily CSF treatment (20 mg/kg) resulted in a stronger reduction in fungal burdens in mice colonized with the wild type strain than in mice colonized with the fen1Δ or fen1-W145* mutants ( Fig 2A ). We also directly compared in vivo fitness of the wild type strain with that of a fen1 mutant by colonizing the mouse gut with a mix of both strains and following each strain’s abundance over seven days, with or without daily CSF treatment. We found that whereas in the absence of CSF treatment, the wild type strain was significantly more fit than either fen1Δ or fen1-W145* strains, in the presence of CSF treatment the two fen1 mutants were more fit, fully taking over the gut population by day 5 ( Fig 2B ). This result was consistent with our initial observation that fen1 mutants were selected in the gut during CSF treatment and confirmed that loss of Fen1 function improved C. glabrata fitness in the presence of CSF.

All gut-evolved strains also contained multiple other coding SNPs relative to the parental strain (DPL1021) and each other, suggesting that these SNPs had arisen during evolution in the mouse gut ( S1 Table ). Among genes previously implicated in echinocandin resistance, GI-63 carried a mutation in the ERG3 gene, which encodes an enzyme involved in the biosynthesis of ergosterol [ 45 ] . Mutations in ERG3 gene can lead to alterations in sterol composition and confer resistance to azoles, as well as echinocandins [ 45 – 47 ] and have also been reported to arise during in vitro evolution of echinocandin resistance [ 36 ].

A. Genetic and echinocandin susceptibility information for 11 gut-evolved mutants. All parental strains used to colonize the mice belonged to the DPL1021 (ATCC90030) background. Some of the parental strains carried the gene deletions shown in the “Genotype” column; “WT” indicates the absence of such gene deletions. These gene deletions had been made as part of our investigation of the roles of the cell wall integrity pathway (slt2Δ and yps1Δ, [ 40 ]) and error-prone DNA replication (rev1Δ and rev3Δ) in the evolution of CSF-resistant mutants in gut-colonizing C. glabrata. B. Diagram of the yeast sphingolipid biosynthesis pathway. FEN1, which encodes a fatty acid elongase, is outlined in red. * = stop codon; CSF = caspofungin; MCF = micafungin; ANF = anidulafungin.

In a previous study, which focused on the role of cell wall integrity pathways in echinocandin tolerance, we examined the evolution of echinocandin resistance during caspofungin (CSF) treatment in the mouse gut C. glabrata colonization model [ 40 ]. Approximately half of the evolved C. glabrata strains showed increased MICs for both CSF and micafungin and had FKS2 hot-spot mutations, but the rest of the strains showed increased MICs for CSF only and lacked mutations in either FKS1 or FKS2 [ 40 ]. We also observed the development of such non-fks mutants in the mouse gut during CSF treatment of mice colonized with C. glabrata strains lacking error-prone DNA polymerases Rev1 and Rev3. To identify the causes of reduced CSF susceptibility in these strains, we sequenced the genomes of seven non-fks mutants with increased CSF MICs, as well as four strains with FKS2 hot-spot mutations that had also evolved in the mouse gut during CSF treatment ( Fig 1A and S1 Table ). Interestingly, we found that six out of 11 sequenced strains–five strains without fks2 mutations and one strain with an fks2 mutation–contained mutations in the FEN1 gene (ELO2 in Saccharomyces cerevisiae) ( Fig 1A ). FEN1 encodes a fatty acid elongase that functions in sphingolipid biosynthesis [ 41 , 42 ] ( Fig 1B ). This finding was consistent with previous reports that mutations in genes of the sphingolipid biosynthesis pathway, including FEN1, reduce C. glabrata and C. albicans sensitivity to CSF [ 43 , 44 ]. Also consistent with those reports, the fen1 mutant strains had increased susceptibility to micafungin (MCF), as well as increased or unaltered susceptibility to anidulafungin (ANF) ( Fig 1A ).

Discussion

In this study, we investigated the evolution of resistance to the widely used echinocandin class drug caspofungin in C. glabrata colonizing the GI tract of immunocompetent mice. We found that, in addition to the well-described mutations in echinocandin target β-glucan synthase, other mutations, predominantly in the FEN1 gene involved in sphingolipid biosynthesis, arose rapidly after the start of caspofungin treatment and increased in abundance in the gut due to their strong fitness advantage over the wild-type strain in the presence of the drug. Our lipidomic, genetic, high-resolution imaging, and MS analyses revealed that the reduced caspofungin sensitivity of fen1 mutant strains is due to elevated intracellular levels of phytosphingosine (PHS), a sphingolipid biosynthesis intermediate, and is associated with reduced caspofungin binding to C. glabrata plasma membrane. Importantly, we also identified several different fen1 mutations in clinical C. glabrata isolates, and our analyses of these mutations showed that they result in loss of protein function and reduced susceptibility to caspofungin. Together, these results identify a new genetic determinant of clinical caspofungin susceptibility in C. glabrata and illuminate the rich dynamics of evolutionary trajectories in gut-colonizing C. glabrata during caspofungin treatment.

Our study illuminates the rapid and complex evolutionary dynamics taking place in the gut C. glabrata reservoir during echinocandin therapy. Both Sanger sequencing of individual colonies and amplicon sequencing of pooled colonies have revealed that mutations in FEN1 emerge and rapidly increase in relative abundance as early as 4 days after the initiation of caspofungin treatment. These results also provide evidence of mixed mutant populations of gut-commensal C. glabrata during caspofungin treatment, where different sub-populations carry different mutations in FEN1, FKS2, and other genes affecting caspofungin sensitivity, and where the relative abundance of these mutant sub-populations fluctuates according to their relative fitness in the presence of caspofungin. Indeed, our amplicon sequencing analysis demonstrated that fen1-W39* and fks2-R1378H alleles arose early during treatment but did not sweep the gut population, indicating that other genetic mutations reducing caspofungin susceptibility had also arisen in the same populations, improving their relative fitness to make them competitive with fen1-W39* or fks2-R1378H. We also provide evidence that in both clinical and mouse gut-evolved C. glabrata strains, fks and fen1 mutations can be acquired sequentially in the same cells (e.g., in GI-297, WM_18.26, and CAS14-5989). Although clinical histories are not always available, the identification of clinical strains containing both mutations in FEN1 and in FKS1 or FKS2 strongly suggests that those mutations had arisen under the same environmental pressure. Together, our results are consistent with a model for evolution of caspofungin resistance in the gut where caspofungin treatment rapidly kills the majority of cells in the gut (which can be seen as the several-log drop in fungal burdens, Fig 2A), but rare drug-tolerant cells survive (Fig 7). These cells may be caspofungin-tolerant either due to pre-existing mutations or to non-genetic mechanisms [56]. Such tolerant cells preferentially proliferate in the gut, and additional mutations further improving fitness in the presence of caspofungin may arise during this proliferation, perhaps stimulated by stress-induced mutagenesis [57,58]. Weak, incremental improvements in fitness can be achieved by a number of genetic alterations (e.g., any loss-of-function mutation in FEN1 or a weak mutation in FKS2) and can therefore occur at a relatively high frequency. The occurrence of these early mutations allows sufficient numbers of C. glabrata cells to survive and proliferate in the gut during caspofungin treatment to eventually develop a rare, clinically echinocandin-resistant mutation in FKS2 (e.g., S663P), which may then take over the population and eventually cause an echinocandin-resistant systemic infection.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 7. Model for evolution of caspofungin (CSF)-resistant C. glabrata in the gut reservoir during CSF treatment. Our results support the model where, upon the initiation of CSF treatment, CSF-sensitive cells are killed but rare CSF-tolerant cells expand in the gut. This CSF-tolerant cells may be due to pre-existing rare mutations or to non-genetic mechanisms. Such CSF-tolerant cells preferentially expand in the gut, and additional mutations with increased fitness during CSF treatment are further selected and expanded. At this stage the gut is a genetically heterogeneous population of cells with increased fitness in the presence of CSF, such as those with mutations in FEN1 or weak mutation in FKS2. Ultimately, this survival and proliferation of the gut C. glabrata reservoir during CSF treatment may culminate in the formation and preferential expansion of a clinically CSF-resistant mutation, such as fks2S663P, which has the highest fitness in the presence of the drug and may eventually cause an echinocandin-resistant systemic infection. https://doi.org/10.1371/journal.ppat.1012521.g007

There are several potential reasons why the gut may provide a permissive environment for the evolution of echinocandin resistance. First, drug concentrations reached in the GI tract during treatment are several times lower than those reached in the plasma and various organs [37]. Second, fungi form prominent biofilms in the GI tract, which are more tolerant to antifungal treatments [59]. Either of these possibilities is consistent with our observation that fen1 mutants have a fitness advantage over wild-type cells in the gut but not in the kidney, where caspofungin concentrations are high enough to eliminate both fen1 and wild-type strains equally well, whereas a fks2 hot-spot mutation with a stronger resistance phenotype is killed less efficiently. Thus, a paradigm emerges wherein in the context of disseminated infection, echinocandin treatment effectively eliminates infecting circulating strains with moderately reduced drug susceptibility (such as fen1), whereas in the context of GI colonization, the same strains survive during treatment, maintaining a continuous viable C. glabrata reservoir in which stronger fks mutations can then emerge (Fig 7), leading to therapy failure. This scenario may be different for the new β-glucan synthase inhibitor ibrexafungerp, which is orally bioavailable and therefore may be present in the GI tract at concentrations high enough to eliminate the gut-colonizing C. glabrata population, but this remains to be experimentally verified.

Several previous studies have reported that mutations in the sphingolipid pathway genes, including FEN1, reduce fungal susceptibility to caspofungin but not to other echinocandins in vitro and that these mutations cause an increase in cellular PHS levels [41,43,49,50]. Our study builds on those results and reveals several key new mechanistic insights. First, we demonstrate that genetic depletion of PHS by deletion of alkaline ceramidase-encoding gene YPC1 sensitizes C. glabrata to caspofungin, strengthening the conclusion that phytosphingosine levels are a key determinant of caspofungin susceptibility. Furthermore, ypc1Δ was fully epistatic to fen1Δ, confirming that increased PHS is the cause of reduced caspofungin sensitivity in the fen1Δ strain. Second, we show that caspofungin association with the cellular surface is significantly reduced in the fen1Δ mutant but is restored in the fen1Δ ypc1Δ mutant, correlating with their reduced and restored caspofungin sensitivity, respectively. This result, as well as our MS analysis of membrane-associated CSF, suggests that the reduced caspofungin efficacy in the fen1Δ mutant is underpinned by its reduced capacity to bind to the plasma membrane due to the membrane’s altered lipid composition (i.e., higher phytosphingosine content), which in turn likely decreases caspofungin ability to inhibit membrane-embedded glucan synthase. A similar hypothesis was proposed by Healey et al. [43], suggesting that the accumulation of long chain bases, such as phytosphingosine, in the plasma membrane reduced the association of caspofungin, but not other echinocandins, with glucan synthase. Further studies are necessary to fully elucidate the effect of sphingolipid pathway mutations on glucan synthase and its interaction with different echinocandin class drugs.

Our results provide one potential explanation for frequent observations of clinical strains showing differential susceptibility to different echinocandin class drugs [15–18,29]. Specifically, based on our results and previous in vitro studies [43], it is likely that strains showing non-susceptibility only to caspofungin are likely to carry mutations in the sphingolipid biosynthesis pathway or, perhaps, mutations in other pathways that cause the accumulation of PHS. We further speculate that other, as yet unknown, pathways differentially affect C. glabrata susceptibility to micafungin and anidulafungin, and that studying the evolution of resistance to these drugs in the mouse gut colonization model can help identify clinically-relevant mutations that reduce C. glabrata susceptibility to one echinocandin while not affecting, or increasing, its susceptibility to the other echinocandins. If such mutants were identified, this information could have important therapeutic implications. For instance, the fact that the fen1 mutations arising in gut-colonizing C. glabrata are highly susceptible to micafungin suggests that, contrary to the conventional wisdom of not combining drugs belonging to the same class, using a combination of caspofungin and micafungin may be of value because it would constrict the pathogen’s evolutionary trajectory towards pan-echinocandin-resistant fks mutations.

Our conclusions about diverse C. glabrata strains coexisting in the gut are highly consistent with genomic analyses of sequential clinical isolates from human patients, which show that a single individual can carry multiple closely related strains, including multiple drug-resistant variants that emerge during drug therapy [24,60]. Together, these observations in the mouse model and in human candidiasis patients underscore the capacity of C. glabrata to undergo micro-evolution in the host and may explain the very high genetic diversity among clinical strains [30, 35]. This capacity for evolution makes C. glabrata a formidable clinical challenge; nevertheless, understanding the mechanisms promoting its evolvability within relevant host niches may help develop effective measures to reduce the evolution of drug-resistant strains.

[END]
---
[1] Url: https://journals.plos.org/plospathogens/article?id=10.1371/journal.ppat.1012521

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.