Limosilactobacillus fermentum Limits Candida glabrata Growth by Ergosterol Depletion

The yeast Candida glabrata, an opportunistic fungal pathogen, and the bacterium Limosilactobacillus fermentum both inhabit the human gastrointestinal and vaginal tract. Lactobacillus species, belonging to the healthy human microbiome, are thought to prevent C. glabrata infections. We investigated the antifungal effect of Limosilactobacillus fermentum on C. glabrata strains quantitively in vitro. ABSTRACT Candida glabrata is a human-associated opportunistic fungal pathogen. It shares its niche with Lactobacillus spp. in the gastrointestinal and vaginal tract. In fact, Lactobacillus species are thought to competitively prevent Candida overgrowth. We investigated the molecular aspects of this antifungal effect by analyzing the interaction of C. glabrata strains with Limosilactobacillus fermentum. From a collection of clinical C. glabrata isolates, we identified strains with different sensitivities to L. fermentum in coculture. We analyzed the variation of their expression pattern to isolate the specific response to L. fermentum. C. glabrata-L. fermentum coculture induced genes associated with ergosterol biosynthesis, weak acid stress, and drug/chemical stress. L. fermentum coculture depleted C. glabrata ergosterol. The reduction of ergosterol was dependent on the Lactobacillus species, even in coculture with different Candida species. We found a similar ergosterol-depleting effect with other lactobacillus strains (Lactobacillus crispatus and Lactobacillus rhamosus) on Candida albicans, Candida tropicalis, and Candida krusei. The addition of ergosterol improved C. glabrata growth in the coculture. Blocking ergosterol synthesis with fluconazole increased the susceptibility against L. fermentum, which was again mitigated by the addition of ergosterol. In accordance, a C. glabrata Δerg11 mutant, defective in ergosterol biosynthesis, was highly sensitive to L. fermentum. In conclusion, our analysis indicates an unexpected direct function of ergosterol for C. glabrata proliferation in coculture with L. fermentum. IMPORTANCE The yeast Candida glabrata, an opportunistic fungal pathogen, and the bacterium Limosilactobacillus fermentum both inhabit the human gastrointestinal and vaginal tract. Lactobacillus species, belonging to the healthy human microbiome, are thought to prevent C. glabrata infections. We investigated the antifungal effect of Limosilactobacillus fermentum on C. glabrata strains quantitively in vitro. The interaction between C. glabrata and L. fermentum evokes an upregulation of genes required for the synthesis of ergosterol, a sterol constituent of the fungal plasma membrane. We found a dramatic reduction of ergosterol in C. glabrata when it was exposed to L. fermentum. This effect extended to other Candida species and other Lactobacillus species. Furthermore, fungal growth was efficiently suppressed by a combination of L. fermentum and fluconazole, an antifungal drug which inhibits ergosterol synthesis. Thus, fungal ergosterol is a key metabolite for the suppression of C. glabrata by L. fermentum.

94 C. glabrata isolates in coculture with L. fermentum (Fig. 1A). We assessed the growth performance of each strain by determining CFU/mL after 20 h of incubation in coculture and in a single-culture setting. We found that the presence of L. fermentum leads to a significant decrease of Candida growth. However, this effect is strain specific, as not all Candida strains were influenced to the same extent. For further analysis we chose the laboratory strain BG2 (29) and 3 clinical isolates (127P, 132P, and 122P) and counted CFU/mL after 24 h (Fig. 1B). The C. glabrata isolates can be grouped into Counted log CFU/mL of Candida glabrata and L. fermentum coculture versus counted CFU/mL of single Candida culture after 24 h; data represent mean log CFU/mL of at least three biological replicates (n = 3 to 5). Asterisks represents a statistical difference between single and coculture conditions (*, P # 0.05). (C) Candida glabrata susceptibility against L. fermentum on MRS medium. Isolates were spotted in serial dilution onto MRS or MRS with 50 mL of L. fermentum culture (OD of 1). (D) Quantitative analysis of growth performance during different conditions and stressors; data represent calculated growth rates after 24 h. Error bars represent the standard error of at least two biological replicates (n = 2 to 4). Asterisks represent a statistical difference between growth rate in SD medium (=control) and growth rate during stress (*, P # 0.05). resistant (BG2 and 127P) and sensitive (122P and 132P) to L. fermentum. The growth performance of isolates against L. fermentum on solid medium revealed the same clustering as in liquid, although some growth reduction could be observed for all isolates (Fig. 1C). Next, we investigated phenotypic traits, such as general stress susceptibility of the isolates, to assess a possible connection to L. fermentum susceptibility (Fig. 1D). We concentrated on conditions associated with L. fermentum stress, for example exposure to lactic and acetic acid or low glucose levels. We quantitatively determined the growth rates of the isolates under these conditions. Strain 132P had the lowest growth rate during various stress tests. The isolates clustered in the same groups as in the L. fermentum sensitivity test, namely, generally lower growth rates (122P and 132P) and generally higher growth rates (BG2 and 127P) (Fig. 1C), except for growth in yeast extract-peptone (YP) medium with 2% lactate or no added carbon source, where all isolates behaved similarly. The same pattern was observed for control conditions (YPdextrose [YPD], Sabouraud dextrose [SD], or MRS medium), which indicates an overall slower growth of 122P and 132P than that of the other two strains. H 2 O 2 stress was the only condition which led to significant growth reduction in the L. fermentum sensitive strains but not in the resistant C. glabrata isolates.
Limosilactobacillus fermentum causes a shift in the Candida glabrata transcriptome. We used transcriptome sequencing (RNA-seq) to investigate how the presence of L. fermentum changes the gene expression pattern of C. glabrata. We compared the expression profiles of a Candida culture with a L. fermentum-Candida coculture after 10 h of incubation. In total, more significantly downregulated differentially expressed genes than upregulated differentially expressed genes ( Fig. 2A) were detected. More differentially expressed genes were found in the two L. fermentum-resistant isolates (BG2 and 127P) than those in the two sensitive isolates (122P and 132P). Principal component analysis visualized the differences between the isolates (Fig. 2B). We conclude that coculture leads to a transcription shift in all C. glabrata isolates and is the main reason for variance (67%) in the samples. Clustering of single-culture profiles points to a similar gene expression pattern under control conditions. PC2 shows the similarity between the different isolates. However, PC2 accounts for only 12% of the total variance. Therefore, the difference of transcriptome profiles between isolates is small compared with the effect caused by different culture conditions.
Looking closer at the differences between the resistant and sensitive isolates, we found a set of genes which are repressed in the Lactobacillus-resistant isolates BG2 and 127P and induced in the Lactobacillus-sensitive isolates (Fig. 2C). Gene ontology (GO) enrichment analysis of the expression data shows genes associated with carbohydrate metabolism and energy reserve ( Fig. 2C; see Table S4 in the supplemental material). Repression of energy reserve metabolism hints that the resistant isolates are not restricting growth and are less bothered by the presence of Lactobacillus than sensitive ones. In the induced gene set, the separation between resistant and sensitive C. glabrata isolates was less clear. In general, we found genes related to steroid metabolism, rRNA metabolism, and iron-sulfur cluster assembly to be upregulated ( Fig. 2C; Table S4).
Interaction with L. fermentum leads to the upregulation of ergosterol synthesis. Clustering using the PathoYeastract approach (30,31) showed that genes regulated by transcription factors were associated with weak acid stress, drug/chemical stress, and iron limitation (Fig. 3A). Weak acid stress was to be expected since L. fermentum is producing acetic and lactic acid. The transcription factor Haa1 is involved in acetic acid stress (32). The gene expression patterns of the Haa1 regulon, as well as acetate-grown C. glabrata, did not reveal significant similarities to our data (see Fig. S1A and B in the supplemental material). Acetate stress may therefore be ruled out as a reason behind the transcriptome shift in coculture.
In yeast, PDR12 is required for weak acid resistance (33)(34)(35) and is the single gene regulated by the weak acid response transcription factor War1. CgPDR12 was upregulated in the Lactobacillus-resistant isolates (Fig. 3B). However, a Dpdr12 knockout mutant was not sensitive to L. fermentum (Fig. 3C). Therefore, a substantial role during Lactobacillus interaction is unlikely for PDR12.
Lactobacilli are able to produce H 2 O 2 and thus may cause oxidative stress. Although the downregulated genes during oxidative stress overlapped with our data set, the upregulated genes were not similar (Fig. S1C). Therefore, the production of H 2 O 2 is unlikely to lead to the observed transcriptomic shift in gene expression.
Response to antifungal drug stress, for example azoles, often involves ergosterol synthesis. Ergosterol is an essential part of fungal cell membrane and is used as a target for antifungals. Interestingly, almost all genes of the ergosterol biosynthesis pathway were found upregulated in all C. glabrata isolates (see Fig. S2 in the supplemental material). Upregulated genes were related to the import and regulation of ergosterol synthesis, such as UPC2A and UPC2B, which are transcription factors responsible for sterol biosynthesis (36), AUS1 responsible for sterol import (37), and HES1 involved in C. glabrata sterol biosynthesis during azole stress (38).
Ergosterol plays a role in response to hypoxia and azole antifungals (39). The coculture setting might cause limiting oxygen availability due to the presence of lactobacilli which leads to an increase of ergosterol biosynthesis. However, we think this process is unlikely because our data do not show an induction of gene sets of the general  Table S4 contains the full list of GO Terms. hypoxia pathways derived from S. cerevisiae (see Fig. S3B in the supplemental material). We further found no differences in the oxygen content of the medium after 10 h of incubation and between coculture conditions and single-culture conditions (Fig.  S3A), thus rejecting a potential hypoxia response in our experiment.
Azole antifungals inhibit cytochrome p450 lanosterol C14a demethylase encoded by ERG11. Deletion of ERG11 prevents ergosterol biosynthesis, leading to an increase of toxic sterols which are fungistatic (8). Fluconazole (FLC) treatment leads to an increase of ERG11 expression in C. glabrata (40). We therefore compared the coculture expression profile data to the gene expression pattern characteristic for fluconazole treatment. Indeed, we found that the expression pattern for fluconazole stress in the strain KUE100 is very similar to the coculture profile (Fig. 4A). C. glabrata isolates are more resistant to fluconazole than C. albicans (9,41). We determined the MIC at 50% growth inhibition (MIC 50 ) of the isolates by measuring the optical density over time and calculating the respective growth rates (Fig. 4B). Fluconazole up to 128 mg/mL reduced the growth rate of all isolates significantly, but it was not fungicidal. The L. fermentum-resistant strains BG2 and 127P had the highest resistance level with an MIC 50 of 32 mg/mL fluconazole, followed by 122P (16 mg/mL) and 132P (8 mg/mL). Thus, the fluconazole resistance profile mirrors that seen with resistance to L. fermentum. Also considering the overlap in gene expression, the findings suggest effective similarities between both conditions for C. glabrata.
C. glabrata ergosterol biosynthesis is not triggered by a simple metabolite. To evaluate the time-dependent effect of the coculture condition, we measured ERG11 expression via quantitative PCR (qPCR) in the BG2 strain (Fig. 4C). ERG11 expression levels peaked after 6 h and was still increased after 10 h of coincubation. After a 24-h incubation, the ERG11 expression level was similar to untreated cultures. This finding suggests an accumulative effect due to an increased biomass of L. fermentum rather than an immediate stress effect. We also explored the following conditions likely caused by L. fermentum exposure: acetic acid, lactic acid, H 2 O 2 , acidic pH (pH 3.5), and glucose depletion (Fig. 4D). We used a 1-h treatment for the immediate effect and a 4-h treatment for the delayed effect. As expected, we found a drop in ERG11 expression level as an immediate effect possibly due to stress-induced transient growth inhibition. After the 4-h treatment, ERG11 expression returned to unstressed levels, suggesting that the tested conditions might not play a decisive role in the upregulation of ergosterol biosynthesis. However, during coculture, a continuous mixture of stressors would be closer to the in vivo situation. To identify a compound or a compound mixture secreted by L. fermentum, we added the culture supernatant and coculture supernatant. A carbon source was added to prevent a starvation response. Both cell-free supernatants did not lead to a significant increase of ERG11. Taken together, we conclude from these experiments that the vicinity of living L. fermentum cells inhibits C. glabrata growth. Ergosterol availability is crucial for C. glabrata growth in coculture. The above results suggested the possibility of a direct role of ergosterol in the L. fermentum C. glabrata coculture system. To test this possibility, we set up a coculture assay on solid medium. L. fermentum cells were spread out onto the plate, and C. glabrata was spotted onto the bacterial lawn. We added ergosterol and different concentrations of fluconazole to the medium (Fig. 5). C. glabrata isolates exposed to 5 mg/mL fluconazole in the presence of L. fermentum showed a slight decrease of growth compared with the control plate with only L. fermentum. Despite all isolates being resistant to 20 mg/ mL fluconazole, a combination of 20 mg/mL fluconazole and L. fermentum abolished the growth of 122P and 132P and significantly reduced the growth of BG2 and 127P (Fig. 5). Fluconazole at a concentration of 10 mg/mL was sufficient to reduce the growth of 122P and 132P which are more sensitive to L. fermentum. Interestingly, the addition of ergosterol reduced the antifungal effect of L. fermentum.
If ergosterol is a key substance for the inhibition of C. glabrata and reduced ergosterol levels due to fluconazole treatment are the cause for more effective inhibition by L. fermentum, an ergosterol depletion by the interruption of its biosynthesis should display a similar phenotype. We thus deleted the ERG11 gene encoding Sterol 14-demethylase from C. glabrata in the BG2 background. We confirmed the knockout mutant with standard methods and show with a chemical analytical approach using a high-pressure liquid chromatography (HPLC)-based method (see Materials and Methods section) that the Derg11 mutant strain does not produce ergosterol (Fig. 6A). We found that this strain is unable to grow in the presence of L. fermentum but remains viable during fluconazole treatment (Fig. 6B). This result is in accordance with previous studies where ERG11 knockout strains as well as a clinical isolate harboring an ERG11 mutation showed resistance toward azoles (42,43). Next, we asked if the ergosterol reduction is specific for the C. glabrata-L. fermentum coculture or occurs in other settings besides the already-described C. albicans-L. rhamnosus GG coculture (23,24). We grew different Candida strains (C. albicans, Candida tropicalis, Candida krusei, and C. glabrata) in single-culture and coculture with different Lactobacilli species (L. rhamnosus and L. crispatus) and determined the ergosterol content of Candida cells (Fig. 6C). Importantly, we observed a significant depletion of fungal ergosterol in L. fermentum coculture across all tested Candida species. L. crispatus, a common vaginal Lactobacillus strain, significantly reduces ergosterol only in C. krusei and C. glabrata, whereas L. rhamnosus, a strain often used in probiotic formulations (25,44), was able to decrease ergosterol content only in C. krusei and C. tropicalis. We conclude that the presence of L. fermentum reduces C. glabrata ergosterol content, which probably triggers the upregulation of steroid metabolism and limits fungal growth, and that this effect is common, albeit to a various degree in other Candida-Lactobacillus coculture settings.

DISCUSSION
The effect of Lactobacillus species on fungi is manifold and dependent on many parameters, such as species, isolate, environment, and time. To describe the general and specific aspects of these interactions, we defined quantitative in vitro coculture conditions, isolated highly and less sensitive C. glabrata strains, and selected as a suitable Lactobacillus strain Limosilactobacillus fermentum (former Lactobacillus fermentum). Previous studies report on the different antifungal capacities of Lactobacillus species against Candida spp. (23,24,45,46). Differences in antifungal activity could be due to the various ability of different lactobacilli isolates to produce metabolic by-products like lactic acid, acetic acid, or H 2 O 2 . However, the production capacity of lactic acid and H 2 O 2 of lactobacilli does not correlate with the effectiveness against Candida species (45). Phenotyping of C. glabrata isolates did not reveal a significant correlation of stress sensitivity and growth inhibition by lactobacilli. Thus, the fungistatic effect of lactobacilli remains unclear.
To narrow down the mechanism of C. glabrata growth inhibition, we defined the transcriptomic response of C. glabrata when it was grown with L. fermentum. The coculture condition lead to a decrease of the adhesion-regulating gene YAK1 and the Yak1-dependent adhesin EPA6 in C. glabrata (see Table S1 in the supplemental material) in accordance with Chew (27). Lactobacillus spp. produce 1-acetyl bcarboline (1-ABC) which is linked to a reduction of filamentation via inhibition of YAK1 in C. albicans (47). Treatment of C. albicans with 1-ABC is not accompanied with a reduction in viability. Therefore, we reasoned that 1-ABC is not the reason for the antifungal effect of L. fermentum against C. glabrata. We found a gene expression pattern pointing at the involvement of an antifungal drug response pathway (Fig. 4A). Some antifungals (azoles) target the fungal plasma membrane containing ergosterol and sphingolipids (48). In accordance with that information, we found an upregulation of ergosterol biosynthesis genes in all isolates (Fig. S2). Previous studies reported that coculture of C. albicans and L. rhamnosus GG leads to a decrease in ergosterol content in the plasma membrane, as well as a downregulation of ergosterol biosynthesis (23,24). Mailänder-Sánchez (24) analyzed the ergosterol content of the C. albicans and L. rhamnosus GG coculture on epithelial cells. In contrast to the findings of Mailänder-Sánchez, in our experimental setting, a coculture with L. rhamnosus did not lead to a significant ergosterol reduction in C. albicans but did in C. tropicalis and C. krusei. The reduction of ergosterol in the Candida cell membrane varied between different Lactobacillus species. Coculture with L. fermentum was the most effective, leading to a dramatic decrease of ergosterol in the fungal cell membrane of C. glabrata, as well as all other tested Candida species (Fig. 6C). We assume that this effect is probably triggering the upregulation of ergosterol synthesis. If ergosterol plays a direct role, a strain with compromised biosynthesis would possibly be hypersensitive to L. fermentum exposure. We generated a Derg11 strain in the BG2 strain background. The Derg11 strain, which cannot produce ergosterol, was indeed unable to grow in the presence of L. fermentum. L. fermentum is able to metabolize cholesterol (20), which is structurally similar to ergosterol. If L. fermentum is able to metabolize ergosterol too, it could lead to the decrease in ergosterol content during coculture.
In C. albicans, lactate as a carbon source leads to a lower expression of CaERG11 and CaERG3 and is accompanied by a reduction of ergosterol content in the plasma membrane (49). Lactobacilli in general produce lactic acid during homolactic fermentation (50). ERG11 encodes a key enzyme of the ergosterol pathway and is regulated according to the demand of ergosterol biosynthesis. We found that CgERG11 was upregulated during coculture but transiently downregulated in the presence of lactic acid after 1 h of treatment and returned to normal after 4 h (Fig. 4D). Therefore, in C. glabrata, lactic acid is unlikely a main player in the L. fermentum-dependent reduction of ergosterol. It is possible that C. albicans, unlike C. glabrata, lacks the response to compensate for Lactobacillus -induced downregulation of ergosterol synthesis and ergosterol content. This finding could explain the higher susceptibility of C. albicans isolates (compared with that of C. glabrata) to the fungistatic effect of Lactobacillus spp.
C. glabrata gene response pattern in coculture with L. fermentum shares similarities to mild FLC stress (51), hinting at a similar response pathway for both conditions. Strains BG2 and 127P had an MIC 50 of 32 mg/mL to fluconazole, whereas 122P and 132P had lower MIC 50 values (16 mg/mL and 8 mg/mL, respectively). BG2 is a clinical isolate known to be unresponsive to FLC treatment (29). FLC-resistant isolates BG2 and 127P were also more resistant to L. fermentum, which further hints to an overlap of fluconazole and L. fermentum response. Previous studies report that a combination of Lactobacillus and azole treatment was more effective at reducing C. albicans and C. glabrata burden than azole treatment alone in in vivo and in vitro (52)(53)(54). We confirm that the combination of fluconazole and L. fermentum decreases the growth of C. glabrata much more efficiently than fluconazole or lactobacilli treatment alone. Lourenço et al. (55) showed that 60 mM acetic acid at a low pH reduces resistance against FLC. The production of acetate was one of the main differences between effective and ineffective lactobacilli; however, concentrations around 4.6 6 0.5 mM were measured (28). Compared with the data of Lourenço et al. (55), it would be too low to exert the combinational effect of acetic acid and FLC toward C. glabrata. In conclusion, we suggest that acetic acid does not play a major role in the combinational effect of FLC and lactobacilli. Taken together, the unavailability of ergosterol, either by fluconazole blocking ergosterol synthesis or by mutation, makes C. glabrata isolates more susceptible toward the effects of L. fermentum.
If ergosterol deprivation is a key determinant of C. glabrata inhibition by L. fermentum, the supplementation of the medium with ergosterol would likely support C. glabrata growth, which was indeed the case. C. glabrata can import sterols under aerobic and anaerobic conditions (56). It was shown that the presence of sterol can mitigate the effect of fluconazole in sterol-synthesis-defective mutants (38). The sterol transporter protein CgAus1 mediates sterol uptake (57). In addition, ERG25 is needed for proper sterol uptake and helps to stabilize sterol-rich lipid domains in the cell membrane (58). However, in an aerobic environment, C. glabrata takes up sterols only during iron limitation conditions (57). It was also shown that iron limitation but not fluconazole stress leads to upregulation of CgAUS1 (57). Furthermore, iron availability was linked to lower Erg11 activity and may subsequently reduce ergosterol content (59). We show here that the presence of L. fermentum leads to the upregulation of CgAUS1 in all four tested isolates, as well as reduced ergosterol content. Additionally, genes regulated by Aft2, a transcription factor responsible for iron homeostasis in S. cerevisiae, were found upregulated under coculture conditions (Fig. 3A) (60). This result could hint at an involvement of the iron homeostasis system during coculture, which enables the import of sterols.
The antifungal effect of Lactobacillus spp. toward C. glabrata is not fully understood. It is currently unclear what leads to a depletion of ergosterol and subsequent growth inhibition of C. glabrata if ergosterol cannot be replenished. The antifungal effect of L. fermentum varies substantially between different C. glabrata isolates. The common transcriptional response of the different isolates clearly revealed steroid biosynthesis as a key factor. However, the strain-specific transcriptional patterns did not allow us to deduce the basis of individual phenotypic differences. A further thorough exploration of the genomes of our Candida isolates could help to understand the genetic basis of the observed phenotypic differences. To our knowledge, we are the first to show that interactions between L. fermentum and C. glabrata lead to a lower ergosterol content in the fungal cell and upregulation of genes for sterol biosynthesis and import. The antifungal effect of L. fermentum can be enhanced by blocking ergosterol biosynthesis with azoles and reduced by ergosterol supplementation (Fig. 7). Finally, we regard the ergosterol depletion effect of Candida by Lactobacillus as strikingly common. These results suggest direct communication between the two species, and it is interesting to note that the bacterium is exploiting the same fungal molecular Achilles heel as conventional antifungal drugs.

MATERIALS AND METHODS
Microbial strains and culture conditions. This study used the clinical isolate BG2 (29) and 93 clinical C. glabrata isolates, as well as C. albicans, C. krusei, and C. tropicalis isolates, which were collected and provided by the Institute of Hygiene and Microbiology at University Hospital St. Pölten. They were propagated in YPD medium at 37°C and 180 rpm. Limosilactobacillus fermentum and L. crispatus were collected and provided by the General Hospital of Vienna. L. rhamnosus (ATCC 53103) was purchased. Standard culture was performed in MRS medium at 37°C and 80 rpm.
Liquid coculture assay. An overnight culture of C. glabrata was diluted to an optical density at 600 nm (OD 600 ) of 0.1 into MRS containing L. fermentum at an OD 600 of 0.05. Single cultures with only Candida were used as the control. Cultures were incubated at 37°C and 80 rpm for 10 h. Serial dilutions  were plated onto YPD supplemented with 50 mg/mL ampicillin. CFUs were counted the next day, and log CFU/mL was calculated. Spotting assay. Overnight cultures of C. glabrata were regrown until an OD 600 of 1. Spots (2 mL) of a serial dilution (1:10) were spotted onto MRS, onto which 50 mL of L. fermentum (OD 600 of 1) was spread previously. Up to 20 mg/ml Fluconazole was added to the MRS medium. For ergosterol supplementation, 4 mg/ml ergosterol (Sigma-Aldrich) was added to the medium. Plates were incubated and inspected every day for up to 6 days.
Library preparation, sequencing, and differential gene expression analysis. Library preparation and sequencing were performed by the Next Generation Sequencing Facility at Vienna BioCenter Core Facilities (VBCF), a member of the Vienna BioCenter (VBC), Austria. PolyA enrichment library preparation and single-read sequencing (50 bp, Illumina HiSeqV4) were obtained of BG2, 122P, 127P, and 132P with and without L. fermentum. At least 2 biological replications of each sample were sequenced and used for subsequent analysis. Demultiplexing was performed by VBCF NGS with bcl2fastq v2.20.0.422. Quality control was performed using FastQC. Reads were aligned against the C. glabrata CBS138 reference genome, which was obtained from the Candida Genome Database (CGD; http://www.candidagenome.org), using Bowtie 2 (64). Quantification of the aligned reads was performed with Rsubread (65) with default settings. Identification of differentially expressed genes (DEGs) between single culture and coculture of each isolate was done using R package DESeq2 (66) with default parameters. Arbitrary cutoff values for DEGs (P # 0.05 and log 2 fold change of $1.5) were chosen. GO term enrichment analysis for the category "metabolic process" was performed using the database of CGD with default settings. Clustering according to transcription factor (TF) was performed using the database of PathoYeastract (http:// pathoyeastract.org). Only documented entries were used for "search by TF." Phenotypic stress test and MIC assay. Overnight cultures of C. glabrata were regrown in YPD, SD full (yeast nitrogen base without amino acids supplemented with ammonium sulfate; BD Difco USA), or MRS medium up to an OD 600 of ;0.5 and diluted 1:10 into the respective medium in a 96-well flat-bottom plate. Stress tests with L-lactic acid (Carl Roth GmbH, Germany), acetic acid (Carl Roth GmbH), H 2 O 2 (Carl Roth GmbH), and at pH 4 and pH 8 were done in SD full medium. For lactic acid and acetic acid, pH was adjusted to pH 4 with HCl; 2% L-lactate (wt/vol) as a carbon source was used in SD full medium, with the pH adjusted to pH 4 with HCl. For every tested condition, C. glabrata isolates were incubated at 37°C in at least triplicates and OD 600 was monitored with an automated set-up (Cytomat42 [Thermo Fisher Scientific, MA, USA], Synergy 95 H1 [Agilent, USA], and Rack Runner 720 [Hamilton Robotics, Germany]). An MIC assay with FLC (0.125 to 256 mg/L) was performed in SD full medium. C. glabrata cells were inoculated with an OD 600 of ;0.05 and incubated at 37°C. The OD 600 was monitored in 2-h intervals for 24 h with a fully automated set-up.
Ergosterol quantification. For the extraction of ergosterol content, the isolates were grown in MRS until an OD 600 of 40 was reached. Coculture samples were separated by centrifugation (2,000 rpm for 5 min). The quantification method was described previously (67). In short, pellets were resuspended with 250 mL 10% (wt/wt) KOH in MeOH, followed by sonication (15 min) and 70°C for 50 min. A total of 50 mL HPLC-grade water and 100 mL N-hexane (Merck, Germany) were added, and the phases were separated by centrifugation. The hydrophobic upper phase was collected followed by a second extraction step. N-Hexane was evaporated at 40°C. When it was completely dry, 100 mL MeOH (Merck, Germany) was added and incubated for 15 min at 40°C. Ergosterol was determined via C 18 reverse phase using HPLC (1200 Series; Agilent Technologies, USA), and detection was performed at 282 nm, a column temperature of 25°C, an isocratic elution of 95% MeOH, and a 5% H 2 O flow rate of 0.3 mL min 21 . The retention time of ergosterol was 3.9 min. Ergosterol content was calculated using a standard curve. Ergosterol content is stated as mg per mL extraction volume.
Statistical analysis. Growth rates for the MIC assay and phenotypic stress tests were calculated using the "growthcurver" package in R (68,69). A heatmap for Fig. 2 was prepared using cluster 3.0 (70) and Java Treeview (71). Clustering for this heatmap was done using complete linkage clustering and correlation as a similarity metric. Other heatmaps were generated using the R package "pheatmap." The genes used for diverse heatmaps are listed in Table S2 in the supplemental material. The principal component analysis was performed and displayed using DESeq2. Dot plots were generated using ggplot2 (72).
Data availability. Raw RNA sequencing data were uploaded to the Gene Expression Omnibus (GEO) under accession number GSE202656.