Comparative Transcriptomics Reveal Possible Mechanisms of Amphotericin B Resistance in Candida auris

ABSTRACT

Candida auris is an emerging multidrug-resistant human fungal pathogen often refractory to treatment by all classes of antifungal drugs. Amphotericin B (AmB) is a fungicidal drug that, despite its toxic side effects, remains a drug of choice for the treatment of drug-resistant fungal infections, including those caused by C. auris. However, the molecular mechanisms underlying AmB resistance are poorly understood. In this study, we present data that suggests membrane lipid alterations and chromatin modifications are critical processes that may contribute to or cause adaptive AmB resistance in clinical C. auris isolates. To determine the plausible cause of increased AmB resistance, we performed RNA-seq of AmB-resistant and sensitive C. auris isolates. Remarkably, AmB-resistant strains show a pronounced enrichment of genes involved in lipid and ergosterol biosynthesis, adhesion, drug transport as well as chromatin remodeling. The transcriptomics data confirm increased adhesion and reduced lipid membrane permeability of AmB-resistant strains compared to the sensitive isolates. The AmB-resistant strains also display hyper-resistance to cell wall perturbing agents, including Congo red, calcofluor white and caffeine. Additionally, we noticed an increased phosphorylation of Mkc1 cell integrity MAP kinase upon AmB treatment. Collectively, these data identify differences in the transcriptional landscapes of AmB-resistant versus AmB-sensitive isolates and provide a framework for the mechanistic understanding of AmB resistance in C. auris.

INTRODUCTION

Candida auris is a multidrug-resistant fungal pathogen that was first reported from the external ear canal infection of a patient in Japan in 2009 (1–3). Within a decade, C. auris has spread around the globe, causing widespread hospital outbreaks of candidemia in some health care settings (4). The Centers for Disease Control and Prevention (CDC) has classified C. auris as an urgent threat to human health due to its clinical and economic impact, high transmissibility, lack of effective antifungal treatments as well as future projections of new infections over the next 10 years (5). According to recent statistics available from the CDC, more than 2000 confirmed cases of C. auris infections have been reported from the United States (6). The spread of C. auris is facilitated through its easy transmission, likely by the contamination of environmental surfaces in hospitals. Thus, C. auris infections present a serious global health threat for several reasons. First, C. auris infections are difficult to treat, since many isolates are multidrug-resistant. C. auris isolates that are resistant to all three major classes of antifungal drugs (azoles, echinocandins and amphotericin B) have also been reported (7, 8). However, the occurrence of pan-resistant isolates is still quite rare. Second, accurate clinical diagnosis of C. auris skin colonization and possible infections is challenging by using traditional diagnostic methods (9–11). Third, this predominant yeast-form haploid pathogen shows persistent colonization of the human body as well as abiotic surfaces in health care environments, thus facilitating the inter- and intrahospital clonal transmission. Finally, recent reports of antifungal pan-resistant clinical isolates (7, 8) highlight the unmet challenges of treating C. auris associated infections and suggest that C. auris carries “pandemic” potential.

The current approved antifungal drugs fall into four main classes: polyenes, azoles, echinocandins and the less often used flucytosine (12). Amphotericin B is a polyene class antifungal drug that was approved in 1958. It has been the first line option for the treatment of cryptococcal meningitis, mucorales as well as infections by many dimorphic fungi (13–16). Amphotericin B is also used as a second line treatment option for invasive aspergillosis (17, 18). Despite its effectiveness, the severe adverse effects of AmB, including nephro- and hepatotoxicity, as well as allergic reactions, have been limiting its long-term use (19). The most widely accepted mode of AmB action suggests that AmB may bind or traps membrane ergosterol, triggering membrane permeability changes by either sequestering ergosterol or by modulating channel function (20, 21). Recent reports suggest a “sterol sponge” mechanism, whereby AmB causes local membrane damages due to spatial ergosterol sequestration (20, 22, 23). However, other data obtained by Raman spectroscopy favor the classical “ion channel” model of AmB and ergosterol binding, leading to membrane permeabilization and leakage (21). Although the precise mechanism of AmB action remains ill-posed and controversial, a general membrane dysfunction appears plausible, suggesting that AmB triggers pore formation in lipid membranes after binding or sequestering ergosterol (20, 22–25). The resulting changes in membrane permeability trigger micropore formation and destroy the electrochemical gradient, causing also an efflux of cellular potassium ions (K⁺) and consequent trigger lysis and cell death (26, 27). Of note, early studies on C. albicans have refuted this theory, suggesting there was no correlation between pore formation and cell death (24, 28). Other modes of AmB action such as inducing apoptosis and oxidative damage have also been proposed (29). Hence, the fungicidal action of AmB is perhaps more complex, and membrane toxicity might not be the sole mechanism of AmB action (28, 29). Therefore, the molecular mechanisms of AmB actions such as how AmB can traverse the rigid carbohydrate cell wall structures, and how it inserts into lipid membranes to form pore structures remain far from being fully understood (21, 25).

Importantly, clinical AmB^R is rare in Candida spp. (26). However, AmB^R has been observed in clinical isolates of several Candida spp., including C. lusitaniae, C. haemulonii species complex, and C. auris (30–33). The extensive clinical use of AmB in some clinical settings are likely to contribute to AmB^R development, but the unrecognized mechanisms C. auris engages for AmB^R highlights a significant gap in mechanistic understanding of C. auris AmB^R.

Here, we use transcriptomics to identify plausible cause(s) of clinical AmB^R in C. auris. Our data show significant enrichment of ergosterol biosynthesis genes, surface adhesion, drug transport as well as chromatin remodeling genes in AmB^R strains compared to susceptible isolates. In addition, the AmB^R strains exhibit increased flocculation and resistance to cell wall perturbation. Furthermore, the AmB^R strains display reduced membrane permeability compared to AmB^S isolates. Overall, our study reveals that changes in membrane lipid permeability is one of the major mechanisms for AmB^R in C. auris.

RESULTS

Antifungal susceptibility of C. auris clinical isolates.

To determine the susceptibility of C. auris strains to antifungal drugs, we performed a broth microdilution assay as described in Clinical and Laboratory Standard Institute (CLSI) document M27-A3 (34). The distribution of MIC values of different antifungal drugs is listed in Table 1. For C. auris, the MIC breakpoints for most antifungal drugs are not well established as yet (6). The current MIC breakpoints are based on those established for related Candida spp. (3, 6). However, the AmB breakpoints are determined based on recent pharmacokinetic/pharmacodynamic studies using murine models of C. auris infection (35). The C. auris isolates with an MIC ≥2.0 are considered resistant to AmB (35). Our results indicate that the C. auris strains 1783/P16, 1184/P15 and 1131/P13 were resistant (MIC = 2 μg/mL) to AmB while the strain 2431/P16 was susceptible (MIC = 0.25 μg/mL) to AmB (Table 1). For convenience, the C. auris clinical isolates 1783/P16, 1184/P15, 1131/P13, and 2431/P16 are mentioned as AmB^R1, AmB^R2, AmB^R3, and AmB^S, respectively, throughout the manuscript.

TABLE 1.

Antifungal susceptibility testing of Candida auris clinical isolates

Strain	Name	Clade	AMB	AFG	CAS	MFG	FLC	ISA	ITC	POS	VRC	5FC
1783/P/16	AmBR1	South Asia	2	0.25	2	0.06	64	1	1	0.25	4	0.125
1184/P/15	AmBR2	South Asia	2	0.125	2	0.06	64	0.06	0.25	0.06	0.25	0.125
1131/P/13	AmBR3	South Asia	2	0.25	0.5	0.25	64	0.25	0.5	0.125	0.125	0.25
2431/P/16	AmBS	South Asia	0.25	0.06	2	0.06	32	0.03	0.25	0.125	0.25	0.125

Transcriptional responses of AmB^R and AmB^S C. auris isolates.

To determine the plausible cause(s) of increased AmB resistance, we performed RNA-seq analysis of logarithmically growing AmB^R in comparison to AmB^S isolates in the Yeast Extract Peptone Dextrose broth medium (YPD; 1% yeast extract, 2% peptone, and 2% dextrose) at 30°C. Differentially expressed genes (DEGs) were defined by a ≥ or ≤ 1.5-fold change (log₂ 0.585), with an adjusted P value cutoff ≤ 0.05 (Table S2). We compared transcript levels of the resistant isolates AmB^R1, AmB^R2, and AmB^R3 to the sensitive isolate AmB^S (Fig. 1, Table S2). The heat map of hierarchical clusters used the Euclidian distance and revealed similar expression pattern between AmB^R2 and AmB^R3 strains compared to AmB^S (Fig. 1A, lanes 1 and 2). However, the expression pattern of AmB^R1 was slightly different from the AmB^R2 and AmB^R3 (Fig. 1A, lane 3). A multivariate principal-component analysis (PCA) revealed similarities between biological replicates as well as differences among the C. auris isolates. The biological replicates from each isolate clustered together, indicating a high level of data correlation, whereas the four isolates clustered separately, with AmB^R1 and AmB^S being separated from AmB^R2 and AmB^R3 by principal component 1 (PC1, 44% of explained variance). Interestingly, both AmB^R1 and AmB^S samples clustered closely, although they showed differential AmB susceptibility profiles (Fig. 1B). A total of 1224 and 737 DEGs were found in AmB^R2 and AmB^R3 strains compared to the AmB^S strain. A relatively low number of 118 DEGs were identified in AmB^R1 cells compared to AmB^S, where 47 genes were induced, and 71 were repressed (Fig. 1C). The smaller number of DEGs in the AmB^R1 strain compared to the AmB^R2 & AmB^R3 isolates may be due to differences in the transcriptional wiring of gene networks among independent clinical isolates. The area-proportional Venn diagram illustrates the striking overlap (approximately 60% of genes) of 433 differentially expressed genes between AmB^R2 or AmB^R3 cells and AmB^S cells (Fig. 1D). Furthermore, an exclusive set of 30 genes were found to be differentially expressed in all three AmB^R strains (Fig. 1D, Table 2). Out of these 30 genes, the expression of 12 genes was increased and the expression of 8 genes was decreased in all AmB resistant isolates, compared to the AmB susceptible isolate. The remaining 10 genes were differentially expressed in opposing directions between AmB^R1 and the other two resistant isolates (AmB^R2 & AmB^R3). For example, the 6 genes that were downregulated in AmB^R1 were upregulated in AmB^R2 and AmB^R3 (Table 2). Overall, these data suggest that there is remarkable heterogeneity among clinical isolates. Additionally, the transcriptional circuitry within C. auris clinical isolates is quite dynamic and differs even among isolates that are deemed to be drug resistant via in vitro MIC assays.

FIG 1.

Transcriptome analysis of C. auris clinical isolates. Gene expression profiles of C. auris isolates grown in YPD were determined using RNA-seq of three biological replicates. (A) Heat map of hierarchical clustering and differentially expressed genes in AmB-resistant isolates compared to sensitive isolate. Lanes 1, 2 and 3 compare AmB^R2, AmB^R3, and AmB^R1 with AmB^S cells, respectively. Log₂-fold change (FC) expression values are color-coded according to the legend on the bottom. (B) Principal-component analysis (PCA) of normalized RNA-seq read counts from three biological replicates per strain displays the level of correlation and the reproducibility among different biological replicates. The two clinical isolates AmB^R1 and AmB^S show high similarity in their gene expression profile. (C) The number of differentially expressed genes along with up (≥1.5-fold) and downregulated (≤ −1.5-fold) genes (FDR ≤ 0.05) are tabulated. Fold change 1.5 = Log₂ FC 0.585. (D) Proportional Venn diagrams depicting the overlap of differentially expressed genes between AmB-resistant isolates relative to the sensitive isolate. (E) The common GO enriched categories (biological process) in AmB^R1, AmB^R2, and AmB^R3 cells compared to AmB^S cells are illustrated in the heatmap. Only significantly enriched (adjusted P ≤ 0.05) GO categories are shown.

TABLE 2.

The list of overlapping genes in all the AmB^R isolates compared to AmB^S isolate

AmBR1 versus AmBS		AmBR2 versus AmBS		AmBR3 versus AmBS		Feature	Gene	C. albicans orthologs
Log2FC	P value	Log2FC	P value	Log2FC	P value
−0.74	0.00	7.79	0.00	5.36	0.00	B9J08_001458	RBT1	RBT1
0.93	0.00	1.94	0.00	1.96	0.00	B9J08_000918	PHR1.1	PHR1
−0.59	0.00	0.73	0.00	1.21	0.00	B9J08_005114	SRR1	SRR1
1.77	0.00	0.76	0.00	1.46	0.00	B9J08_004027	WOR1	WOR1
−0.62	0.00	1.63	0.00	0.86	0.00	B9J08_001362	AGP3	AGP3
−0.64	0.00	0.79	0.00	0.91	0.00	B9J08_004365	SFC1	SFC1
2.47	0.00	3.84	0.00	1.23	0.00	B9J08_004100	HYR3.1	HYR3
0.61	0.00	0.99	0.01	0.71	0.04	B9J08_005544	SEC26.2	SEC26
0.87	0.00	1.93	0.00	2.44	0.00	B9J08_000266	B9J08_000266	orf19.4666
1.16	0.00	1.08	0.00	0.78	0.00	B9J08_000447	B9J08_000447	orf19.6650
−1.07	0.00	1.44	0.00	1.18	0.00	B9J08_001339	B9J08_001339	orf19.6487
2.08	0.00	1.50	0.00	1.54	0.00	B9J08_002118	B9J08_002118
0.85	0.00	2.66	0.00	2.81	0.00	B9J08_003109	B9J08_003109	orf19.2638
−0.63	0.00	0.70	0.00	0.68	0.00	B9J08_004053	B9J08_004053	orf19.4445
−0.62	0.00	1.13	0.00	−0.67	0.00	B9J08_004260	B9J08_004260
1.19	0.00	1.03	0.00	2.20	0.00	B9J08_005063	B9J08_005063
0.65	0.00	2.89	0.00	1.97	0.00	B9J08_004971	DDR48	DDR48
1.68	0.00	2.49	0.00	3.02	0.00	B9J08_000363	NCE103	NCE103
0.62	0.00	0.99	0.00	0.71	0.00	B9J08_000811	NDL1	orf19.6148
−0.70	0.00	−2.73	0.00	−1.57	0.00	B9J08_002582	ALS4	ALS4
−0.87	0.00	−0.94	0.00	−0.90	0.00	B9J08_004839	PST3.2	PST3
0.88	0.00	−0.82	0.00	1.00	0.00	B9J08_001531	IFF4.1	IFF4
−1.22	0.00	−1.06	0.00	−1.16	0.00	B9J08_004108	MLT1.2	MLT1
0.65	0.00	−1.03	0.00	−0.94	0.00	B9J08_000261	ERG1	ERG1
−0.79	0.00	−2.09	0.00	−0.83	0.00	B9J08_002239	HGT2.1	HGT2
0.74	0.00	−2.58	0.00	−0.65	0.00	B9J08_002583	B9J08_002583	orf19.6520
−0.89	0.00	−3.92	0.00	−1.89	0.00	B9J08_004173	LIP1.4	LIP1
−0.98	0.00	−3.03	0.00	−2.82	0.00	B9J08_005554	RTA2.2	RTA2
−0.86	0.00	−1.52	0.00	−0.73	0.01	B9J08_002240	B9J08_002240	orf19.1664
−0.65	0.00	−1.17	0.00	−0.98	0.00	B9J08_002245	B9J08_002245	orf19.3139

AmB^R C. auris strains show enrichment of adhesion, chromatin remodeling, drug transport, and sterol biosynthesis genes.

The Gene ontology (GO) enrichment analysis of DEGs identified by RNA-seq analysis revealed a significant enrichment of GO biological process in all AmB^R strains, including pathogenesis, transmembrane transport, ergosterol biosynthesis, and filamentous growth (Fig. 1E and Fig. S1). The significantly enriched strain-specific GO processes (Fig. S1A) and commonly shared GO categories (Fig. S1B) between different AmB^R strains are depicted in Fig. S1. To identify larger patterns in differential gene expression and to obtain an overall insight into the differential transcriptional landscapes between the AmB^R versus AmB^S isolates, we performed GO annotation and manually inspected DEGs. We found genes encoding cell wall adhesins, drug transporters, chromatin remodelers, and the constituents of ergosterol biosynthesis to be significantly differentially expressed in AmB^R isolates compared to the AmB^S isolate (Fig. 2). Approximately 73% of (30 out of 41) genes encoding proteins involved in cell adhesion were differentially expressed across three AmB^R isolates (Fig. 2A, Table S3). Of these, PHR1.1, (B9J08_000918; C. albicans homolog PHR1 [CaPHR1]), and HYR3.1 (B9J08_004100; CaHYR3) were highly upregulated. The GPI-anchored cell wall protein IFF4.1 (B9J08_001531; CaIFF4) was upregulated in AmB^R1 and AmB^R2, and the other ortholog IFF4.2 (B9J08_004451; CaIFF4) was upregulated in AmB^R2 and AmB^R3. In addition, AmB^R3 cells showed approximately 13-fold and 2-fold increased expression of ALS1 (B9J08_004498), and ALS3 (B9J08_004112), respectively (Fig. 2A). Interestingly, ALS4 (B9J08_002582; CaALS4) was downregulated in all the AmB^R isolates (Fig. 2A). Differential expression of ALS adhesins was indeed confirmed by qPCR (Fig. 3A). C. albicans ALS gene expression such as ALS1 and ALS3 is controlled by the bHLH transcription factor Efg1 (36–38). Since divergent upstream signaling cascades can converge at their downstream transcriptional regulators such as Efg1 and Cph1 to modulate morphogenesis and virulence in C. albicans (39–41), we quantified expression of EFG1 and CPH1 in C. auris isolates. The qPCR results revealed a significantly increased expression of EFG1 and CPH1 in AmB^R2 and AmB^R3 strains, compared to the AmB^S isolate. We noticed that there was no change in the expression of EFG1 in the AmB^R1 isolate. However, the CPH1 expression was significantly reduced in the AmB^R1 strain, compared to the AmB^S strain (Fig. S3B and C). To assess whether differential expression of adhesin genes have biological effects, we performed crystal violet assays to quantify the adhesion of C. auris isolates to plastic (Fig. S3A). Our data clearly indicate that AmB^R2 and AmB^R3 strains were hyperadherent to plastic compared to the AmB^S isolate (Fig. S3A).

FIG 2.

Differential regulation of gene sets implicated in adhesion, chromatin remodeling, drug transport and sterol biosynthesis. (A–D) Heat maps displaying differentially regulated genes in adhesion, chromatin remodeling, drug transport and sterol biosynthesis, respectively. In all the panels, lanes 1, 2 and 3 compare AmB^R2, AmB^R3, and AmB^R1 with AmB^S cells, respectively. Gene names were adopted from the respective C. albicans orthologs and unidentified orthologues are stated as C. auris feature name (ex: B9J08_004154, B8441 annotation). Log₂-FC expression values are color-coded according to the legend on the bottom of each heat map.

FIG 3.

RT-PCR analysis of C. auris adhesins, KATs and drug transporters. (A–C) Quantitative RT-PCR (qPCR) analysis of differentially expressed genes in YPD-grown AmB^R1, AmB^R2, AmB^R3 and AmB^S cells. Data are shown as mean of relative expression to the reference gene GAPDH from three independent experiments (±SEM, *, P < 0.05; **, P < 0.01; ***, P < 0.0005; ****, P < 0.0001). The mRNA expression levels of ALS gene family (A), chromatin remodelers (lysine acetyl transferases) (B), and drug transporters (C) are depicted.

According to the CGD annotations, a total of 113 and 48 genes from C. albicans and C. auris have been directly annotated to the term chromatin remodeling through computational analysis (42). Nearly 38% of the chromatin remodelers (18 out of 48 genes) were differentially expressed in both AmB^R2 and AmB^R3 compared to the AmB^S isolate (Fig. 2B). We observed two genes encoding histone H1 (HHO1; B9J08_000855), and H3 (HHT2; B9J08_003368), and the H3K56 deacetylase HST3 (B9J08_002059) (43) to be upregulated in both AmB^R2 and AmB^R3 isolates. Interestingly, lysyl acetyltransferases (KATs) including GCN5 (B9J08_005133), HAT2 (B9J08_005442), ELP3 (B9J08_001912), HPA2 (B9J08_002171), and SPT10 (B9J08_002077) (44) were differentially regulated in both AmB^R2 and AmB^R3 isolates. These transcriptomics data were validated by qPCR (Fig. 3B). However, none of these genes were significantly altered in AmB^R1 compared to the AmB^S isolate (Fig. 3B, Table S4).

Recent studies have indicated high levels of intrinsic expression of membrane transporters in AmB^R C. auris strain B11210 (45). Interestingly, changes in the expression of major drug transporters was not observed in strain B11210 (45). However, we observed differential expression of several efflux pumps from the ATP-binding cassette (ABC) and major facilitator superfamily (MFS) transporter superfamilies (46, 47) in AmB^R strains. These included the MFS transporters, HOL1 (B9J08_004040; CaHOL1), HOL4 (B9J08_004662; CaHOL4), MDR1.2 (B9J08_004113; CaMDR1) upregulated in AmR² and AmR³ (Fig. 2C, Table S5). In addition, we also found induction of the ABC transporter CDR1 (B9J08_000164; CaCDR1) in AmB^R3 compared to AmB^S isolate. Of note, MFS antiporters encoded by QDR genes, including QDR1 (B9J08_005492; CaQDR1), QDR2 (B9J08_002663; CaQDR2), QDR3.2 (B9J08_005403; CaQDR3) were significantly repressed, whereas QDR3.1 (B9J08_005172; CaQDR3) was induced in AmB^R2 isolate compared to the AmB^S isolate (Fig. 2C). Again, quantitative real-time qPCR confirmed the differential expression of both ABC drug efflux pumps (CDR1) and MFS pumps (MDR1, QDR1, QDR3) (Fig. 3C).

Genes belonging to the ergosterol biosynthesis ERG1, ERG2, ERG4, ERG5, ERG25, and ERG27, were differentially expressed and validated by qPCR (Fig. 2D and 4, Table S6). However, there were notable differences in the expression profiles of ergosterol biosynthesis genes in the individual strains used in this study. For example, ERG1 was mildly upregulated in AmB^R1 but it was significantly downregulated in the AmB^R2 and AmB^R3 strains (Fig. 4). Similarly, the ERG2 was downregulated in AmB^R2 and AmB^R3 but there was no change in its expression in AmB^R1 (Fig. 4). We noticed a decrease in the expression of ERG6 and ERG13 in the AmB^R3 strain relative to AmB^S (Fig. 4). We hypothesize that this difference in expression of ergosterol biosynthetic pathway genes among AmB resistant strains may be due to heterogeneity in clinical isolates. An approximately 1.5-to 2-fold increased expression of PDR16 encoding a phosphatidylinositol transfer protein and increases of sterol uptake gene SUT1 was observed in the AmB^R3 strain (Fig. 2D). Furthermore, among several identified transcription factors (TFs), WOR1 (B9J08_004027; CaWOR1) was upregulated in all AmB^R isolates compared to the AmB^S isolate (Fig. S2A, Table S7). Interestingly, we observed a significant overlap (22 TFs out of 77 identified) in the expression of various TFs between AmB^R2 and AmB^R3 isolates (Fig. S2A).

FIG 4.

AmB-sensitive (AmB^S) isolate shows differential expression of ergosterol biosynthesis pathway. Transcript levels of several ergosterol biosynthetic pathway components were measured with qPCR. Data depict mean values of relative expression to the reference gene GAPDH from three independent experiments (±SEM, *, P < 0.05; **, P < 0.01; ***, P < 0.0005; ****, P < 0.0001).

Previous studies have shown the importance of two-component response regulators SRR1, SSK1 and downstream mitogen-activated protein kinase (MAPK) pathways in antifungal drug resistance and cell wall homeostasis in C. albicans and C. auris (48–50). Several genes involved in the MAPK pathways such as Hog1-mediated stress response signaling, Mkc1-mediated cell integrity pathway, and Cek1/Cek2-mediated starvation or filamentation were differentially regulated in AmB^R strains (Fig. S2B, Table S8). Taken together, the marked differences in the transcriptional landscapes of AmB^R versus AmB^S C. auris strains highlight the importance of several gene families that may contribute either alone or through genetic interactions to drive adaptive AmB^R.

Activation of MAPK pathways in AmB^R C. auris isolates.

To gain more insight into MAPK signaling events in AmB^R, we determined the activation of Mkc1 and Hog1 MAP kinases. We performed immunoblotting of cell extracts from AmB^R and AmB^S strains (Fig. 5), demonstrating elevated basal levels of Mkc1 and Hog1 phosphorylation in the AmB^S isolate compared to the AmB^R strains. However, AmB treatment further increased phosphorylation of Mkc1 in all AmB^R strains. Interestingly, the opposite effect of impaired phosphorylation signal was observed in the AmB^S strain. There were minor changes in Hog1 phosphorylation after treatment with AmB in AmB^R1 and AmB^R2 C. auris strains. Therefore, based on these observations, we conclude that the Mkc1 cell integrity signaling pathway contributes to the regulation of AmB^R in C. auris.

FIG 5.

AmB-resistant isolates show differential activation of both Mkc1 and Hog1 MAPK pathways. (A) Logarithmically growing Candida auris cultures were treated with 500 mg/mL amphotericin B for 15 min and washed once with ice-cold water. Whole-cell extracts for immunoblotting were prepared by the TCA method. Extracts corresponding to 0.5 OD₆₀₀ were fractionated by 12% SDS-PAGE and subjected to immunoblotting as indicated, using commercially available antibodies for the activated phosphorylated MAP kinases Mkc1-P (Phospho-p44/42 MAPK (Erk1/2) (Cell Signaling) and Hog1-P (Phospho-p38, Cell Signaling). Reprobing the blots with the PSTAIR antibodies (Sigma) recognizing Cdc28 served as a loading control. Protein bands were visualized using an Odyssey CLx scanner (Li-Cor). (B–C). Densitometric analysis of Mkc1-P (B) and Hog1-P (C) blots using image studio software (LI-COR). Striped column bars indicate drug (AmB) treatment conditions. Data are expressed as fold change normalized to the PSTAIR (Cdc28) loading control. Data represent mean values from three independent biological samples (± SEM, *, P ≤ 0.05; **, P ≤ 0.005).

AmB^R strains show increased flocculation and elongated cell morphologies.

To determine the biological effect of differential expression of several genes related to cell wall function in AmB^R isolates, we subjected all C. auris isolates to extensive phenotypic profiling to test for morphogenesis, and susceptibility to cell wall stress agents (Fig. 6 and Fig. S5). In our previous study, AmB^R2 isolate (1184/P15) showed flocculation and a multicellular aggregation phenotype (50). We first examined the flocculation phenotypes of strains growing in the YPD broth at 30°C. Consistent with a previous report, AmB^R2 cells showed flocculation and sedimented at a higher rate. In addition, we observed that AmB^R3 also displayed a pronounced flocculation phenotype, although it was absent in AmB^R1 and AmB^S strains (Fig. 6A). Microscopic examination of flocculating AmB^R2 and AmB^R3cells revealed approximately 10% and 5% of cells showing elongated morphologies, respectively (Fig. 6B). Furthermore, AmB^R2 and AmB^R3 isolates showed more clumps and aggregates (data not shown).

FIG 6.

Phenotypic profiling of Candida auris clinical isolates. (A) Flocculation of AmB resistant and sensitive clinical isolates was tested. All the samples were vortexed and photographed after 2 and 10 min. Arrows indicate the floccules and precipitation in the suspension. (B) Representative confocal microscopy image showing C. auris cells stained with calcofluor white (CFW) to detect cell wall chitin. Logarithmically growing cells were fixed, washed and stained with CFW 1 mg/mL. Differential Interference Contrast (DIC) and UV light images (UV) of the same cells were scanned at ×63 magnification with 2× digital zoom. White color arrowheads indicate the elongated cells, and the approximate percentage of these cells are depicted in the top right corner of the image, wherever is indicated. Scale bars represent 5 μm.

AmB^R and AmB^S C. auris display differential susceptibilities to cell wall stress.

To interpret the effects of changes in the transcriptional landscapes of AmB^R versus AmB^S C. auris strains, we tested the growth of C. auris isolates on solid and liquid media supplemented with specific antifungal drugs as well as various stress agents, including caffeine, CFW, Congo red (CR), and sodium dodecyl sulfate (SDS) (Fig. S5). Phenotypic characterization of the AmB^S isolate revealed increased susceptibility to antifungal drug AmB, cell wall stressors such as caffeine, CFW, SDS, and CR, although sensitivity to the antifungal drugs fluconazole (FLC) and caspofungin (CAS) was not significantly altered (Fig. 5A and B). In addition, the AmB^R3 isolate showed slightly increased sensitivity to CFW (Fig. S5). Interestingly, the AmB^R1, AmB^R2, and AmB^S isolates exhibited poor growth at higher temperature 42°C (thermal stress) (Fig. S5B). Of note, these phenotypic data strongly support and confirm the transcriptomic data sets.

Clinical AmB^R C. auris isolates display reduced membrane permeability.

Based on our RNA-seq data (Fig. 1) showing enrichment of genes involved in membrane transport and ergosterol biosynthesis, we hypothesized that AmB^R C. auris clinical strains have distinct membrane lipid features that may affect nonprotein mediated lipid membrane permeability. Therefore, we measured the kinetics of fluorescein diacetate (FDA) uptake by C. auris strains. FDA is a lipophilic nonfluorescent precursor dye whose diffusional uptake kinetics is solely determined by membrane lipid fluidity, which in turn determines membrane permeability (50, 51). Indeed, the data show markedly reduced membrane permeability of all AmB^R strains in comparison to the AmB^S strain (Fig. 7). The AmB^S cells showed an increased linear FDA accumulation, resulting in 10 to 30-fold higher permeability than the AmB^R cells (Fig. 7). These data suggest that membrane lipid permeability changes provide a major barrier element for diffusional uptake of antifungal drugs. Collectively, based on our RNA-seq, and membrane permeability data, we propose that dynamic alterations of membrane lipid composition may constitute a major factor in the development of AmB^R in C. auris.

FIG 7.

AmB-sensitive (AmB^S) isolate shows altered membrane permeability. The kinetics of fluorescence-based fluorescein diacetate (FDA) uptake was measured. FDA uptake was allowed with continuous shaking; fluorescence readings were taken every 5 min for 60 min. The slope was calculated using GraphPad Prism and is depicted on the right side. Data represent the mean fluorescence intensity from three biological replicates (±SEM; ****, P < 0.0001).

DISCUSSION

The polyene class drug amphotericin B (AmB) is one of the first antifungals with fungicidal action (52). The antifungal spectrum of AmB is quite broad since it shows a significant amount of activity against a wide variety of human fungal pathogens. The mode of action of AmB is quite different compared to other antifungals. AmB, unlike most antifungal drugs, does not target a specific enzyme. Instead, it exerts its antifungal activity by directly interacting with ergosterol present in the fungal cell membrane (20). This may be one of the reasons why AmB resistance is quite rare in fungal pathogens. Another plausible explanation for this paradox is that AmB resistance often comes with severe fitness costs in C. albicans (53). It is not clear whether AmB resistance in C. auris comes with fitness tradeoffs. Although a recent study with C. auris clade II strain B11220 demonstrated that by conducting in vitro microevolution experiments resistance to AmB comes with fitness costs (54). Additionally, mutations in ERG3 and ERG11 genes were shown to be associated with cross-resistance to AmB and fluconazole in C. auris strain B11220 (54). More recently, mutations in ERG6 were identified to be associated with an increased AmB resistance in C. auris (55). Mutations in the ergosterol biosynthesis pathway have also been associated with increased AmB resistance in C. albicans and C. glabrata (56, 57).

The Candida auris strains cluster into 5 different clades (58). These clades differ by several thousand SNPs. However, within each clade the isolates are reported to be clonal. Among all clades, the South Asian clade contains highest percentage of AmB^R as well as MDR isolates (59). Therefore, this study is focused on AmB^R mechanisms in South Asian clade. Furthermore, C. auris frequently displays high resistance to AmB. In some cases, 30 to 50% of C. auris isolates, within a clonal cluster, exhibit AmB^R (3, 59). However, the precise molecular mode of action of AmB has remained a matter of dispute. Thus, we aimed here to use transcriptomics to better understand the mechanisms underlying AmB^R in a defined set of clinical C. auris patient isolates. We identify the differences in the transcriptional landscapes of AmB^R strains in comparison the AmB^S C. auris strains. The present study identifies several gene families that were differentially expressed in AmB^R C. auris isolates. These include genes required for cell wall function (PHR1, HYR3.1, IFF4.1, and IFF4.2) adhesion (ALS1, ALS3 and ALS4), lipid and ergosterol biosynthesis (ERG1, ERG2, ERG4, ERG5, ERG25, and ERG27), as well as for chromatin remodeling (HST3, GCN5, HAT2, ELP3 and SPT10). These findings suggest that AmB resistance in C. auris is influenced by changes in cell wall, lipid and ergosterol biosynthesis. Importantly, our results correlate with a recently published report showing enrichment of membrane transport and ergosterol biosynthesis genes in AmB^R C. auris strain B11210 (45). In the present study, we used three distinct AmB-resistant (AmB^R1, AmB^R2 and AmB^R3) isolates and a single AmB susceptible (AmB^S) isolate from the South Asian clade. These isolates were selected solely on the basis of in vitro MIC values (Table 1). It is important to note that these isolates do not cover the breadth of large C. auris intraspecies diversity. Therefore, to gain an in-depth understanding of AmB resistance in C. auris, future transcriptional analysis experiments involving more isolates from the South Asian as well as other clades may be required.

An important finding from this current study is that we noted that the transcriptomes of AmB^R2 and AmB^R3 were quite similar. However, the transcriptome of the AmB^R1 isolate was relatively different compared to the other two resistant strains (AmB^R2 and AmB^R3). This is all the more interesting because the phenotypes (e.g., membrane lipid permeability, Mkc1 MAPK phosphorylation, etc.) of all AmB^R strains are similar. The observed differences in the transcriptomes of AmB-resistant isolates may be due to alterations in the transcriptional wiring of gene networks, thus leading to heterogeneity in clinical strains. We hypothesize that the heterogeneity among C. auris clinical isolates may add to the genetic diversity. Heterogeneity in clinical isolates of Candida albicans, C. auris, and Aspergillus fumigatus is frequently encountered and has also been described by several previous studies (60–63).

Since the mechanism of AmB^R is not well understood, several hypotheses attempt to explain the increased antifungal resistance traits in AmB^R isolates. One of the most widely accepted causes for AmB^R has been both qualitative and quantitative changes in the lipid as well as the reduction in total ergosterol content of the fungal cell membrane (56, 57). Indeed, our results show that AmB^R C. auris strains display reduced membrane lipid permeability. The transcriptional profile of AmB^R strains also show significant differential expression of ergosterol biosynthesis genes compared to the AmB^S strain. These data are in line with published reports, suggesting an important role of total ergosterol content and ergosterol biosynthetic genes for AmB^R in Candida lusitaniae and Aspergillus spp (64, 65).

However, changes in membrane lipid permeability and ergosterol content may not be the only drivers of AmB^R in C. auris. Indeed, our results indicate that other mechanisms such as chromatin modifications and MAPK signaling may also contribute to MDR in C. auris. The gene expression profile of AmB^R strains compared to the AmB^S strains revealed changes in the expression of lysine acetyl transferases (KATs) and lysine deacetylases (KDACs). These results fit well with the current understanding of importance of fungal KATs and KDACs in the regulation of antifungal drug resistance (38, 43, 66, 67). Fungal KATs and KDACs work primarily by modifying both histone and nonhistone target genes. For example, the pleiotropic KAT Gcn5 act on histones H3 and H4 as well as regulate cell mediated MAPK signaling pathway (38, 44). Additionally, fungal KATs/KDACs work in close cooperation with dedicated transcriptional regulators, thereby forming a dual-layer network of chromatin-mediated transcriptional control (67–69).

Fungal pathogens often sense signals for adaptation to environmental stress, regulation of virulence factors, as well as resistance to antifungal drugs via two-component signaling and MAPK pathways (50, 70). In this regard, activation of Hog1 and Mkc1 MAPK signaling pathways via phosphorylation, in response to AmB and caspofungin (CAS) has been demonstrated in C. auris (50). The present study demonstrates activation of Mkc1 via phosphorylation in AmB^R strains when exposed to AmB. Based on these features we hypothesize that regulation of drug resistance via MAPK signaling pathways is critical in C. auris.

Taken together, our findings offer significant new insights into possible mechanisms driving clinical AmB^R resistance in C. auris strains. These studies also provide a framework for further studies to mechanistically dissect AmB resistance in C. auris.

MATERIALS AND METHODS

Candida auris strains and growth conditions.

Candida auris clinical isolates were grown in YPD broth (1% yeast extract, 2% peptone, and 2% dextrose) at 30°C with shaking at 200 rpm. Logarithmic-phase cells were obtained by growing overnight cultures in fresh YPD medium for 4 h at 30°C. Two percent agar was added to the plates. The C. auris strains used in this study are listed in Table 1. The primers used for the qPCR are listed in Table S1 in the supplemental information.

Antifungal susceptibility testing of C. auris strains.

Antifungal susceptibility of C. auris strains was performed according to the guidelines described in CLSI document M27-A3 (34). The antifungals tested were amphotericin B (AmB, Sigma, St. Louis, MO, USA), anidulafungin (AFG, Pfizer, Groton, CT, USA), caspofungin (CAS, Merck, Whitehouse Station, NJ, USA), micafungin (MFG, Astellas, Toyama, Japan), fluconazole (FLC, Pfizer), isavuconazole (ISA, Basilea Pharmaceutica, Basel, Switzerland), itraconazole (ITC, Lee Pharma, Hyderabad, India), posaconazole (POS, Merck), voriconazole (VRC, Pfizer), and 5-flucytosine (Sigma). All antifungals were dissolved in DMSO. RPMI 1640 medium with glutamine without bicarbonate pH 7 (Sigma) was buffered with 0.165 M MOPS (Sigma). The MIC endpoint for AmB was defined as the lowest drug concentration that caused 100% growth inhibition compared to the drug-free control at 35⁰C. Whereas for all other drugs, the MIC endpoints were defined as the lowest drug concentration that caused ≥ 50% reduction in fungal growth respective with the control wells.

Growth and phenotypic profiling.

To study the effect of in vitro stressors and antifungal drugs, the C. auris strains were grown in YPD broth overnight at 30°C. From an overnight culture, cells corresponding to an optical density at 600 nm (OD₆₀₀) of 0.1 were inoculated into fresh YPD broth with or without caspofungin (200 ng/mL), amphotericin B (500 ng/mL), calcofluor white (50 μg/mL), caffeine (50 mM), and SDS (0.1%). Absorbance was recorded in a multimode microplate reader (SpectraMax iD5, Molecular Devices, San Jose, CA, USA) at regular intervals for a period of 24 h, and the OD₆₀₀ values were plotted versus time. Additionally, the phenotypic characterization of C. auris strains was done via serial dilution spotting assays on YPD agar plates. Equal volumes (3 μL) of 10-fold serial dilutions of logarithmically growing C. auris strains were spotted onto YPD agar plates containing different stress agents such as temperature stress (42°C), hydrogen peroxide (H₂O₂, 7.5 mM), caffeine (50 mM), SDS (0.1%), calcofluor white (CFW, 50 μg/mL), caspofungin (CAS, 100), amphotericin B (AMB, 2 μg/mL), and fluconazole (FLC, 128 μg/mL). Colony growth was assessed after 48 h.

Microscopy.

To stain cell wall chitin with Calcofluor White (CFW; Fluorescent Brightener 28, Sigma), 1 mL of logarithmically growing cells were fixed in 4% p-formaldehyde for 2 h. Fixed cells were washed and stained with CFW 1 mg/mL for 5 min. UV light images (UV) and Differential Interference Contrast (DIC) of the same cells are scanned at ×63 magnification with Leica TCS SP8 confocal microscope. Approximately 100 to 200 cells were counted, and elongated and pseudo-hypha-like cells were represented in percentage. Scale bar = 5 μM.

Flocculation assay.

Flocculation was determined by growing the C. auris strains to the late exponential growth phase in YPD broth at 30°C as previously reported (50). Equal OD₆₀₀ cells were transferred to separate culture tubes. The culture tubes were vigorously vortexed and allowed to settle for 2 and 10 min. Images were recorded after 2 min and 10 min.

Adherence assay.

Adherence on polystyrene-coated plates was measured by crystal violet staining as described previously (38). Briefly, samples containing 2 OD₆₀₀ Candida cells in YPD were loaded into a flat-bottomed 96-well microtiter plate (CytoOne) and incubated for 4 h at 30°C. The culture medium was aspirated, cells washed once with PBS to remove nonadherent Candida, and fixed for 15 min with 100% methanol. Plate wells were allowed to dry at room temperature, and cells were stained for 5 min with 200 μL of 1% crystal violet (vol/vol). Then, cells were washed gently once with water, followed by the addition of 200 μL of 33% acetic acid, and absorbance was measured at 570 nm. Data are expressed as the mean of absorbance of Candida strains from four independent biological replicates.

Western blot analysis.

Logarithmically growing C. auris clinical isolates were treated with or without AMB (1xMIC of AmB for each C. auris strain) for 10 min. After that, cultures were washed once with ice-cold water, and whole-cell extracts were prepared by the trichloroacetic acid (TCA) method as described previously (50). Briefly,3 to 5 OD₆₀₀ of cells were harvested, resuspended in 1 mL water, and incubated on ice with 150 μL YEX lysis buffer (1.85 M NaOH, 7.5% β‐mercaptoethanol) for 10 min. Proteins were precipitated by adding 200 μL of 50% trichloroacetic acid and incubated on ice for 10 min. Samples were centrifuged at 13 000 × g, and the pellet was resuspended in sample buffer (40 mM Tris–HCl pH 6.8, 8 M urea, 5% SDS, 0.1 mM EDTA, 0.1 M tris base, 0.1 g/liter bromophenol blue, and 1% β‐mercaptoethanol). Extracts corresponding to 0.5 OD₆₀₀ were fractionated by 12% SDS-PAGE and blotted for proteins as indicated. Signals from the same whole-cell extracts were detected using antibodies for active phosphorylated MAP kinases. The commercial antibodies recognized phosphorylated Mkc1-P and Cek1-P (phospho-p44/42 MAPK [Erk1/2]; Cell Signaling Technology) and Hog1-P (phospho-p38; Cell Signaling Technology). Reprobing with PSTAIR antibody (Sigma) recognizing Cdc28 (B9J08_002497) served as a loading control. Protein bands on the nitrocellulose membrane were visualized using an Odyssey CLX scanner (Li-Cor). Quantification of the protein band intensity was performed by using Image Studio software (Li-Cor). The intensity ratios of phosphorylated versus loading control were used to generate a bar graph of the densitometry in GraphPad Prism software.

Transcriptional profiling using RNA-sequencing (RNA-seq).

For RNA-seq analysis, overnight-grown C. auris cultures were inoculated into YPD (initial OD₆₀₀ of 0.1) and grown at 30°C for 4h. Total RNA was purified using the Genejet RNA purification kit (Thermo Scientific). Quality of RNA was assessed on a Bioanalyzer using the RNA6000 Nano chip (Agilent), mRNA was enriched using oligonucleotide(dT) beads (NEB), and subsequently, double-stranded cDNA libraries were generated by using the NEBNext Ultra Directional RNA Library Prep kit for Illumina (NEB) according to the manufacturer’s instructions. The qualified libraries were subjected to Illumina NovaSeq 6000 sequencing platform with a 150 bp paired-read length at the Novogene (Novogene, USA) sequencing facility. Three biological replicates for each strain were sequenced. The raw RNA-seq data are deposited to the Gene Expression Omnibus (GEO) under the accession number GSE190920.

Raw reads from RNA-seq libraries were filtered to remove the adaptor sequences, low-quality reads (the quality value of over 50% bases of the read is ≤ 20), and the reads containing N > 0.1% (N represents base that could not be determined). After filtering, C. auris B8441 reference genome was downloaded from Candida Genome Database (http://www.candidagenome.org) and mapped to the clean reads using HISAT2 (71). HTSeq (72) was used to count the read numbers mapped for each gene. The FPKM (Fragments Per Kilobase of transcript sequence per Millions base pairs sequenced) of each gene was calculated based on the gene read counts mapped. Read count obtained from gene expression analysis was used to do differential expression analysis using the DESeq2 R package (73), while the significant criterion is false discovery rate (padj) <0.05 and log₂-fold change ≥0.585.

For further downstream analysis, C. albicans orthologs were retrieved from the Candida Genome Database (http://www.candidagenome.org). Several C. albicans genes were mapped to more than one gene feature of C. auris. To avoid ambiguity, the duplicate genes were given an extra number chronologically. For instance, B9J08_003981, and B9J08_004113 features were best matched to C. albicans MDR1 and were named MDR1.1, and MDR1.2, respectively. Gene ontology (GO) annotations were used the GO term finder tool (http://www.candidagenome.org/cgi-bin/GO/goTermFinder), and the online bioinformatics tool Fungifun2 (https://elbe.hki-jena.de/fungifun/) (74). Heat map of hierarchical clustering and differentially expressed genes was generated using data mining tool Orange3 (75). Venn diagrams were generated using BioVenn (http://www.biovenn.nl/index.php) (76). The RNA-seq analysis results are presented Tables S2-S8.

Fluorescein diacetate (FDA) uptake assay.

The kinetics of FDA uptake was performed as described previously (50). Briefly, logarithmically growing C. auris strains were harvested at about 0.5 OD. Cells were resuspended and washed twice in 1 mL of FDA buffer (50 mM HEPES, pH 7.0, and 0.5 mM 2-deoxy-d-glucose) before supplementing with 50 nM FDA. A 200-μL volume of cell mixture with or without FDA was added to an optical-bottom 96-well black plate. The kinetics of FDA uptake was recorded every 5 min for 30 reads or until saturation was reached with simultaneous shaking on the H1 Synergy plate reader with excitation and emission wavelengths of 485 and 535 nm, respectively. Data represent mean fluorescence intensity over time. The slope was calculated using GraphPad Prism software.

Quantitative real-time PCR (qPCR) analysis.

Total RNA was purified using Genejet RNA purification kit (Thermo Scientific). cDNA synthesis and qPCR analysis was done as described previously (38). The efficiency-corrected ΔΔCt method was used to quantify mRNA expression levels of a target gene transcript in comparison to a reference gene transcript (77). The mRNA of the gene associated with GAPDH (TDH3: B9J08_001227) was used as a reference gene. GraphPad Prism software was used to perform statistical analyses of independent biological replicates (n = 3) of indicated strains.

Statistical analysis.

All data in this study are presented as mean ± standard error of mean (SEM), unless otherwise stated. All graphs were plotted using the GraphPad Prism software. All statistical analysis was performed by a two-sided Student's t test, unless otherwise stated. The number of biological replicates is stated in each figure legend. The significance P values are indicated in the relevant figure panels and in figure legends.