Heteroresistance to Fluconazole Is a Continuously Distributed Phenotype among Candida glabrata Clinical Strains Associated with In Vivo Persistence

ABSTRACT Candida glabrata causes persistent infections in patients treated with fluconazole and often acquires resistance following exposure to the drug. Here we found that clinical strains of C. glabrata exhibit cell-to-cell variation in drug response (heteroresistance). We used population analysis profiling (PAP) to assess fluconazole heteroresistance (FLCHR) and to ask if it is a binary trait or a continuous phenotype. Thirty (57.6%) of 52 fluconazole-sensitive clinical C. glabrata isolates met accepted dichotomous criteria for FLCHR. However, quantitative grading of FLCHR by using the area under the PAP curve (AUC) revealed a continuous distribution across a wide range of values, suggesting that all isolates exhibit some degree of heteroresistance. The AUC correlated with rhodamine 6G efflux and was associated with upregulation of the CDR1 and PDH1 genes, encoding ATP-binding cassette (ABC) transmembrane transporters, implying that HetR populations exhibit higher levels of drug efflux. Highly FLCHR C. glabrata was recovered more frequently than nonheteroresistant C. glabrata from hematogenously infected immunocompetent mice following treatment with high-dose fluconazole (45.8% versus 15%, P = 0.029). Phylogenetic analysis revealed some phenotypic clustering but also variations in FLCHR within clonal groups, suggesting both genetic and epigenetic determinants of heteroresistance. Collectively, these results establish heteroresistance to fluconazole as a graded phenotype associated with ABC transporter upregulation and fluconazole efflux. Heteroresistance may explain the propensity of C. glabrata for persistent infection and the emergence of breakthrough resistance to fluconazole.

, and in the pathogenic yeast Cryptococcus neoformans (14). When heteroresistant strains are serially cultured on media containing inhibitory concentrations of an antimicrobial drug, each generation exhibits variable drug responses but the nonsusceptible subpopulation gradually expands and fully resistant colonies will ultimately emerge. Population analysis profiling (PAP) assays are considered the gold standard for determining HR (13). In a wider context, HR is a manifestation of bet hedging, a mechanism whereby genetically identical cells express different phenotypic profiles, thus increasing the probability of survival during stressful fluctuations of environmental conditions (15). Potential implications of HR include treatment failure, relapse, and the establishment of persistent chronic infection (13). HR is generally not detected in standard broth microdilution susceptibility assays, and failure to detect HR may result in misclassification of nonsusceptible strains as susceptible by clinical microbiology laboratories. However, the lack of a standardized definition of HR complicates attempts to determine the clinical importance of this phenomenon (13).
We explored fluconazole HR (FLC HR ) in C. glabrata by performing PAP of clinical isolates. We then tested the gene expression and function of ABC drug transporters in C. glabrata strains exhibiting different degrees of FLC HR . Finally, we determined the effect of HR on in vivo survival during fluconazole treatment. Our results demonstrate that fluconazole HR is a continuously distributed phenotype in C. glabrata clinical isolates. The area under the PAP curve provides a quantitative measurement of FLC HR and correlates with ATP-dependent efflux transporter activity. Importantly, FLC HR is not detected by routine susceptibility testing methods and could be an important factor underlying treatment failure.

RESULTS
Drug susceptibility population distribution analysis. The fluconazole susceptibility characteristics of 52 C. glabrata clinical isolates and three reference strains (median fluconazole MIC of 2 mg/liter; range, 0.25 to 16 mg/liter; see Table S1 in the supplemental material) were determined by PAP. PAP assays were analyzed using the heterogeneity range (HR) method, which categorizes strains in a dichotomous fashion as fluconazole heteroresistant (FLC HR ) or nonheteroresistant (FLC N ), and by using the area under the curve (AUC) ratio (AUCR), which provides a continuous measurement of heterogeneity (Fig. 1). AUCRs ranged from 0.00012 to 13.74 (median 0.46; interquartile range, 0.10 to 1.12; Fig. 2A). Correlating the AUCR with fluconazole doseresponse heterogeneity showed a partition between AUCRs for HRs of Ͻ16 and Ն16. Specifically, the median AUCR was 0.09 (range, 0.00012 to 0.41) for an HR of Ͻ16 and 0.96 (0.25 to 13.74) for an HR of Ն16 (P Ͻ 0.0001; Fig. 2B). An AUCR breakpoint of 0.43 categorized 28 (93%) of 30 strains with an HR of Ն16 as FLC HR and 24 (96%) of 25 strains with an HR of Ͻ16 as FLC N . However, for each HR value of Ն16, the AUCRs varied up to 30-fold: 0.25 to 5.85 for an HR of 16, 0.42 to 13.74 for an HR of 32, and 1.66 to 8.32 for an HR of 64 (Fig. 2). This indicates that rather than being a bimodally distributed phenotype, FLC HR is a continuously distributed trait in C. glabrata. In all, 30 (57.6%) of 52 clinical C. glabrata isolates were FLC HR according to an HR breakpoint of Ն16. Population fluconazole susceptibility profiles were stable for each strain, with minimal variation in fluconazole AUCRs after Ͼ10 passages on fluconazole-free medium (data not shown). The fluconazole MIC was slightly but significantly higher for FLC HR strains than for FLC N strains (MIC 50 [interquartile range], 2 [1 to 4] g/ml versus 1 [1 to 2] g/ml; P ϭ 0.02).
To determine the prevalence of HR to echinocandins, we performed caspofungin and anidulafungin PAP assays of all of our C. glabrata isolates. All of the isolates tested had a heterogeneity range of 2 to 4 for echinocandins, indicating no significant HR to these drugs. There was no difference in echinocandin HR among isolates that were FLC N , FLC HR , and fluconazole resistant (Fig. 3). All of our C. glabrata isolates grew on 2% (wt/vol) glycerol as the sole carbon source, indicating the absence of mitochondrial mutations leading to respiration-defective (petite) mutants (16).
Rhodamine 6G efflux and accumulation correlate with HR. We measured the glucose-dependent efflux activity of ABC-type transporters in representative FLC HR and FLC N C. glabrata strains with the rhodamine 6G assay (17). As expected of ABC transporters, rhodamine 6G efflux, determined by relative fluorescence in supernatants, occurred rapidly after the addition of glucose to yeast cell suspensions and was absent from glucose-free controls (Fig. 4). Glucose-dependent rhodamine 6G efflux was significantly higher in FLC HR strains than in FLC N strains (P Ͻ 0.001; Fig. 4). However, rhodamine 6G efflux activity varied markedly within the FLC N and FLC HR categories, with no clear breakpoint between FLC N and FLC HR strains, as defined by heterogeneity range criteria. Moreover, rhodamine 6G efflux correlated positively and linearly with the degree of FLC HR , expressed as the AUCR (r 2 ϭ 0.68, P ϭ 0.0021; Fig. 4).
Evaluation of rhodamine 6G retention in yeast cells by fluorescence microscopy allows visual assessment of efflux activity within a cell population. There was cell-to-cell variation in rhodamine 6G accumulation within both FLC HR and FLC N populations (Fig. 5). However, the proportion of cells stained with rhodamine 6G following a 25-min incubation in phosphate-buffered saline (PBS) with 8 mM glucose was lower for the FLC HR strain than for the FLC N strain. Accumulation of rhodamine 6G was almost absent from FLC R cells and was lower than that in both FLC HR and FLC N strains (Fig. 5).
Taken together, these results indicate that the fluconazole AUCR of a given C. glabrata strain reflects the overall ABC transporter efflux activity of the cell population, which in turn is an expression of the proportion of cells with increased efflux activity.
Individual ABC transporter genes are upregulated in a binary manner in FLC HR strains. We compared the ACT1normalized expression of the three azole efflux transporter genes (CDR1, PDH1, SNQ2) and the ABC transport regulator gene PDR1 among FLC N (n ϭ 4), FLC HR (n ϭ 6), and FLC R (n ϭ 2) C. glabrata strains (Fig. 6). As expected, in FLC R strains, CDR1 expression was dramatically higher than in both FLC N and FLC HR strains (mean expression, 47.6-fold Ϯ 19.8-fold; P Ͻ 0.0001) (Fig. 6). Modest but significant increases in CDR1 expression levels were observed in four of six FLC HR strains, whereas none of four FLC N strains had increased CDR1 expression (mean expression, 2.32-fold Ϯ 0.04-fold and 1.43-fold Ϯ 0.19-fold, respectively; P ϭ 0.028). In addition, PDH1 expression was significantly increased in five of six FLC HR strains versus none of the FLC N strains (mean expression, 2.63-fold Ϯ 0.56-fold versus 1.29fold Ϯ 0.37-fold; P ϭ 0.015). In sum, four of six FLC HR strains upregulated both CDR1 and PDH1 and an additional FLC HR strain upregulated PDH1 alone (Fig. 6). In contrast, there was no significant difference in the expression of SNQ2 and PDR1 between FLC N and FLC HR strains, suggesting that transporters other than Cdr1 and Pdh1 (Snq2 and the Pdr1 transcription factor) do not play a major role in FLC HR .
Unlike fluconazole efflux, expression of CDR1 and PDH1 was not linearly related to the fluconazole AUCR. The fold expression values of both genes were distributed bimodally among FLC HR and FLC N strains, suggesting dichotomous "on-off" states of gene expression (see Fig. S1 in the supplemental material). Both FLC HR strains that did not upregulate CDR1 had low rhodamine 6G efflux activity (see Fig. S1). Upregulation of CDR1 and PDH1 was associated with high values of rhodamine 6G efflux, but the difference from that in strains not upregulating these genes did not reach statistical significance (P ϭ 0.1; see Fig. S1).
Preexposure to fluconazole (8 or 64 mg/liter) resulted in significant upregulation of CDR1 expression in two FLC HR strains: Cg2268, which upregulated CDR1 at the baseline (2.8-fold increase in CDR1 expression following preexposure; P Ͻ 0.0001), and Cg1922, which did not upregulate CDR1 at the baseline (50% increase in expression; P ϭ 0.008; Fig. 6). There was no significant effect of fluconazole exposure on the CDR1 expression of two other FLC HR strains. In contrast, PDH1 expression in all FLC HR strains preexposed to fluconazole was reduced to 38 to 70% of the baseline values (Fig. 6). The expression levels of both SNQ2 and PDR1 following preexposure to fluconazole remained well below the 2-fold threshold for all FLC HR strains, indicating that CDR1 upregulation in FLC HR strains is not mediated through altered Pdr1 levels (18). Rhodamine 6G efflux increased in all C. glabrata strains after incubation with fluconazole, but the increase was more rapid and pronounced in strain Cg2268 than in FLC N strains Cg304 and Cg1775 (65% increase [P Ͻ0.0001] versus 26 to 51% increase [P ϭ 0.01]; Fig. 6C). Thus, upregulation of CDR1 by fluconazole preexposure was associated with enhanced rhodamine 6G efflux.
Sequencing of PDR1 in six FLC HR strains and four FLC N strains revealed seven nonsynonymous mutations in the coding sequence (see Table S2 in the supplemental material). PDR1 mutations were detected in three of six FLC HR strains and two of four FLC N strains. These mutations were not associated with a gain-offunction phenotype and did not corresponded to previously described PDR1 gain-of-function substitutions (19)(20)(21).
HR is associated with persistence of viable C. glabrata cells in mouse kidney tissue during fluconazole treatment. To determine whether FLC HR correlates with failure of fluconazole to eliminate C. glabrata from visceral tissue, we assessed the viable fungal burdens (in numbers of CFU per gram of tissue) in mouse kidneys 7 days after intravenous inoculation. High-dose fluconazole treatment (100 mg/kg/day) significantly reduced the tissue fungal burdens of FLC N and FLC HR C. glabrata-infected mice (P Ͻ 0.0001) and, as expected, had no effect on FLC R C. glabratainfected mouse tissue fungal burdens (Fig. 7). Individual mice with persistent C. glabrata infection after 7 days of fluconazole treatment could be detected in both the FLC N and FLC HR infection groups. Persistent infection manifested itself as high tissue fungal burdens approaching those observed in untreated mice (2.0 ϫ 10 5 Ϯ 7 ϫ 10 4 and 8.6 ϫ 10 5 Ϯ 1.6 ϫ 10 5 CFU/g, respectively; Fig. 7). Importantly, the frequency with which persistent infection was detected was higher in fluconazole-treated mice infected with a highly FLC HR strain (11 of 24 mice, 45.8%) than in fluconazole-treated mice infected with an FLC N strain (3 of 20 mice, 15%) (P ϭ 0.029), despite both strains having fluconazole MICs in the susceptible range.
FLC HR is variable within phylogenetic clusters. Comparison of the PAP phenotypes and phylogenetic relationships determined from intergenic spacer (IGS) sequences showed clusters of C. glabrata strains with similar phenotypes, as well as variable AUCRs, among closely related strains (Fig. 8). Two clusters (A and C) were composed of closely related FLC HR strains. Interestingly, both clusters also contained FLC R strains (one in cluster A and two in cluster C). The six fully fluconazole-resistant strains included in this analysis did not cluster with each other; four of the FLC R strains were found within clusters shared by FLC HR strains, and two strains were singletons (Fig. 8).
Population heat mapping of C. glabrata strains according to AUCRs showed that closely related C. glabrata strains often have divergent AUCRs (clusters A, B, C, F, and G; Fig. 8). In sum, phylogenetic analysis shows both phenotypic clustering and dynamic changes in the fluconazole susceptibility patterns of related C. glabrata strains.

DISCUSSION
Stochastic cell-to-cell variation in gene expression within genetically identical populations is a postulated mechanism for microbial survival under fluctuating environmental conditions, also referred to as bet hedging (22). HR is a specific manifestation of bet hedging as it pertains to antimicrobial resistance (11)(12)(13)23). However, fundamental questions remain regarding HR. First, a uniform definition of HR has not been agreed upon (13). More importantly, it remains unclear whether HR is a binary characteristic or a continuously distributed phenotype and whether similar definitions should apply to prokaryotic bacteria and eukaryotic fungi (15). Second, the molecular underpinnings of HR remain unclear. Finally, the importance of HR as a mechanism of microbial persistence and treatment failure in vivo has yet to be determined (11)(12)(13). Here, we show that FLC HR is a continuously distributed phenotype in C. glabrata, a major human fungal pathogen that excels in the ability to become fully resistant following exposure to this drug (4). Quantitative measurement of HR by using the fluconazole AUC of the population analysis plot (AUCR) provided an indirect assessment of the proportion of resistant cells within the total population of each strain, as evidenced by the linear correlation between the AUCR and ABC-type efflux activity. Importantly, in vivo persistence during treatment with high-dose fluconazole was significantly associated with FL-C HR , providing evidence that HR may be a cause of treatment failure and relapse observed in the clinical setting (24).
Using a binary definition of HR suggested in a recent review (13,25), 57.6% (30/52) of the fluconazole-susceptible C. glabrata isolates included in this study met the criteria for FLC HR . However, the continuous distribution of the AUCR among FLC HR and FLC N strains (Fig. 2) and the correlation of this index with rhodamine 6G efflux indicate that FLC HR is more appropriately treated as a graded phenotype. This observation is consistent with the work of Levy et al. (15), who described a continuum of growth rates within the cell population of Saccharomyces cerevisiae, a nonpathogenic yeast closely related to C. glabrata. Those researchers hypothesized that in yeast, a combination of stochastic and deterministic factors dictate a spectrum of inheritable epigenetic states that confer different levels of fitness (15). In contrast, a bimodal distribution of persister and nonpersister cells has been described in bacterial populations (22). Our findings extend the graded nature of yeast cell states to the interstrain population level.
In the majority of FLC HR isolates, we detected a low level of ABC transmembrane transporter gene CDR1 and PDH1 upregulation (4). These expression levels were much lower than those seen in FLC R isolates (i.e.,~2-fold versus 50-fold), likely because in a heteroresistant population, only about 1 in 10 3 cells exhibits a resistance phenotype, thus diluting the observed increase in transporter transcripts. Moreover, none of the FLC HR C. glabrata isolates showed evidence of a gain-of-function mutation in PDR1, which has been reported to cause full resistance to fluconazole (19). Consistent with this, no significant increase in SNQ1 expression, whose transcription is also regulated by Pdr1p, was detected.
In contrast to rhodamine 6G efflux, CDR1 and PDH1 expression was increased to similar levels in FLC HR strains, irrespective of their AUCRs (Fig. 7). Thus, the FLC HR phenotype seems to represent the incremental effects of multiple binary genetic switches, whose sum produces a spectrum of HR states.
The clinical significance of HR remains largely unclear for the bacteria in which it has been described (11). Here, we determined the correlation between high-level HR and persistent infection during a simulated course of high-dose fluconazole treatment by using a mouse model of hematogenously disseminated candidiasis. A significantly larger proportion of mice infected with an FLC HR isolate than of mice infected with an FLC N strain had high kidney fungal burdens after 7 days of fluconazole treatment. Interestingly, some of the FLC HR strains did not show a kidney fungal burden at day 7 and a kidney fungal burden was observed in mice infected with the FLC N strain, albeit less frequently. This finding is consistent with our in vitro results and supports the notion that some degree of HR is present in most C. glabrata isolates.
Internal diversity in population drug susceptibility may arise from genetic, epigenetic, or nongenetic mechanisms. We analyzed genetic relatedness to gain insight into the evolution and dynamics of the FLC HR phenotype. We found phenotypic clustering of some FLC HR and FLC N strains, as well as significant variability of AUCRs among some genetically related C. glabrata isolates. These relationships suggest that FLC HR represents an interplay of genetic and epigenetic mechanisms (Fig. 8).
Resistance of C. glabrata to fluconazole varies widely between isolates from different geographical regions, as well as within those regions (2,3). We found high rates of FLC HR among C. glabrata clinical isolates collected in Israel and the United Kingdom. Together with our phylogenetic data, this geographic pattern is consistent with the circulation of multiple C. glabrata clades with different propensities to evolve full azole resistance. The degree to which regional differences in azole use drives this process requires further study.
Given the epidemiological and clinical importance of C. glabrata, we suggest that HR may be a driver of azole resistance, treatment failure, and unresolved infection. Of note, surveillance efforts have focused solely on azole resistance and do not account for the HR phenotype, which may be much more prevalent. This discrepancy may explain the inconsistent correlation observed between fluconazole MICs and clinical outcomes of treating C. glabrata infection with this drug (26). Interestingly, we did not detect HR to echinocandins in any of the clinical C. glabrata isolates, consistent with the better correlation between echinocandin MICs and the responses to treatment (8,27).
It has been postulated that because of the fungistatic nature of azoles, Candida cells exposed to these drugs enter an arrested state in which stress-induced genetic instability promotes the develop- was measured in the presence of 8 mM glucose (ϩ glucose) and under glucose starvation conditions (Ϫ glucose) to differentiate ABC transmembrane transporter specific activity. White bars, FLC N strains; dark bars, FLC HR strains. In aggregate, FLC HR had significantly greater ABC-type efflux activity. Yet, as with the AUCR, rhodamine 6G efflux was distributed continuously rather than bimodally between FLC HR and FLC N strains. RLU, relative luminescence units. (B) Correlation between rhodamine 6G efflux activity and degree of FLC HR , expressed as FLC-AUCR. Empty circles, FLC N strains; black squares, FLC HR strains. ment of drug resistance and breakthrough infection (28). In this context, enhanced in vivo persistence of FLC HR C. glabrata during fluconazole treatment, as observed in the mouse model, creates optimal conditions for the emergence of stress-induced mutations conferring azole resistance. The lack of significant HR to echinocandins may be related to the rapid fungicidal effect of glucansynthase inhibition, which does not allow time for stress-induced genetic instability. It should be noted, however, that of the 52 clinical C. glabrata strains in this study, only 8 were recovered from patients treated with an azole during the preceding year (see Table S1 in the supplemental material). Hence, other types of environmental stress were probably involved in the generation of FLC HR strains. ABC transporters have broad substrate specificity, allowing them to export a wide range of cytotoxic molecules, including organic acids, sterols, and H ϩ ions (29). FLC HR may thus represent part of a more general stress-induced phenotype associated with the upregulated expression of ABC transporter genes.
Mechanisms of HR remain poorly understood at the molecular and single-cell levels (23). Our findings establish that FLC HR appears to be frequent in clinical C. glabrata isolates, affecting more than half of the strains tested. Quantitative PAP analysis revealed a continuous range of FLC HR rather than a categorical phenotype, suggesting that multiple determinants are involved. Additional studies should be undertaken to identify the range of genetic and/or physiological determinants that contribute to FLC HR and to elucidate the significance of the FLC HR phenotype in clinical practice.

Ethics statement.
All animal experiments were approved by the Tel Aviv Sourasky Medical Center (TASMC) Institutional Animal Care and Use Ethics Committee (IACUC) under protocol b3703-11-037 and conform to the Israeli Animal Experimentation Law and U.S. Public Health Service Policy on Humane Care and Use of Laboratory Animals. Clinical data extraction was approved by the TASMC Ethics Committee under protocol 0217-11-TLV in adherence to the Israeli Ministry of Health guidelines for clinical trials with human subjects.
Candida strains. C. glabrata strains were obtained from clinical samples collected at the TASMC microbiology laboratory (48 strains) and at the University of Exeter, United Kingdom (10 strains). Each strain was specific to a single patient. Strains were identified to the species level with the Vitek 2 system and the YST ID card (bioMérieux, Marcy l'Etoile, France) and by sequencing of internal transcribed spacer (ITS) regions ITS1 and ITS2 (30). All of our C. glabrata isolates (38 bloodstream and 14 mucosal isolates) were determined to be susceptible to fluconazole, voriconazole, and itraconazole by broth microdilution according to CLSI methods outlined in document M27-A3 (31). C. glabrata strains CBS15126, CBS138, and BG2 were used as fluconazole-susceptible reference strains. C. glabrata clinical isolates Cg1708, Cg191, and Cg326 served as fluconazole-resistant controls.
Drug susceptibility PAP. Although there are no formal standards for determining HR, PAP is generally considered the most robust method (13). We analyzed the population distribution of Candida isolates for susceptibility to fluconazole, caspofungin, and anidulafungin by a plating dilution method (32). Yeast extract agar glucose (YAG; yeast extract at 5 g/liter; glucose at 10 g/liter; agar at 15 g/liter; 1 M MgCl 2 at 10 ml/liter; trace elements at 1 ml/liter) was prepared with nine different concentra-

FIG 5
Visualization of intracellular rhodamine 6G accumulation. Rhodamine 6G accumulation in cells was detected by fluorescence microscopy. Panels: A to C, FLC N isolate Cg1775; D to F, FLC HR isolate Cg2268; G and H, FLC R isolate Cg1708; A, D, and G, light microscopy; B, E, and H, fluorescence microscopy (rhodamine filter); C, F, and I, composite images. All images were captured at ϫ400 magnification by using manually fixed exposure values. DIC, differential interference contrast.
tions of fluconazole (log 2 serial dilutions ranging from 128 to 0.5 mg/ liter), caspofungin (8 to 0.03 mg/liter), and anidulafungin (8 to 0.03 mg/ liter). Twenty milliliters of each YAG-drug concentration or a drug-free YAG control was poured into 20-cm culture plates and allowed to cool. Candida strains were passaged once on Sabouraud's agar to ensure purity. Two or three colonies were suspended in sterile saline, and hemocytometer counts were used to adjust the density to 10 6 yeast cells/ml. Four log 10 dilutions of yeast cell suspension were prepared, corresponding to 10 5 to 10 2 CFU/ml. Each YAG-drug plate was divided into quadrants, and each quadrant was inoculated with six 5-l drops of a different yeast cell suspension concentration and marked accordingly. Plates were incubated at 30°C for 24 h, and the optimal yeast concentration at which nonconfluent colonies could be enumerated was used to count CFU and to calculate the number of viable cells per milliliter. For each Candida strain, the viable cell density (in CFU per milliliter) was plotted over the drug concentration (13). To quantitate FLC HR , we calculated the area under the population distribution curve (AUC) (32). In order to increase the sensitivity of this method to detect variations in Candida growth at high fluconazole concentrations, we analyzed the AUCs for fluconazole concentrations ranging from the MIC to 128 mg/liter (Fig. 1). To enhance robustness across different experiments, the AUC was normalized by dividing it by the same value for a well-characterized FLC HR C. glabrata strain (Cg1646), yielding an AUCR.
Bacterial HR is commonly defined in binary terms as a minimal fold difference between the lowest drug concentration associated with maximal growth inhibition and the highest noninhibitory concentration (25),  in four FLC N isolates (white bars), six FLC HR isolates (gray bars), and two FLC R isolates (black bars). Each dashed line marks the arbitrary 2.0-fold threshold of upregulation. *, P Ͻ 0.05; **, P Ͻ 0.01 (for the comparison of gene expression with the mean expression of four FLC N isolates). Note that the y axis is on a different scale for each gene. (B) Induction of efflux transport-associated genes following a 3.5-h incubation in drug-free medium or medium containing fluconazole (FLC) at 8 or 64 g/ml. *, P Ͻ 0.05; **, P Ͻ 0.01 (for relative gene expression induction following drug exposure). Expression was normalized to ACT1 and then compared to levels of expression in the same strain that was not exposed to the drug. PDR1 expression never reached the 2-fold threshold for upregulation. Each data point represents the mean Ϯ the standard deviation of three experiments. (C) Rhodamine 6G efflux of C. glabrata strains Cg304, Cg1775 (FLC N ), and Cg2268 (FLC HR ) after incubation with fluconazole (8 g/ml, black bars) or a saline control (gray bars). *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.0001. referred to here as the heterogeneity range (Fig. 1). To determine a meaningful fold difference breakpoint for HR, we analyzed the correlation between the AUCR and the heterogeneity of drug response on PAP.
Rhodamine 6G efflux and accumulation analyses. To assess ABCtype drug transporter activity, we determined the glucose-induced efflux and cellular retention of rhodamine 6G as described previously (17,33). C. glabrata isolates were grown to log phase in liquid YAG at 35°C. Yeast cells were collected by centrifugation, and 10 9 cells were transferred to 20 ml of fresh YAG and incubated at 27°C for an additional 2 h. For drug exposure experiments, cells were incubated in PBS with fluconazole at 8 g/ml or in drug-free PBS at 30°C in a shaking incubator for 3 h. Next, yeast cells were collected by centrifugation and washed twice in PBS, and 10 ml of PBS containing 15 M rhodamine 6G without glucose was added to the pellets. Suspensions were vortexed and incubated at 27°C for 90 min to allow rhodamine 6G uptake under carbon source depleted conditions. Cells were collected by centrifugation, washed twice in PBS, and suspended in 750 l of PBS in microcentrifuge tubes. To start rhodamine 6G efflux, 250 l of PBS with 8 mM glucose of was added. Control tubes were prepared with glucose-free PBS. Tubes were removed after 5, 15, and 25 min of incubation at 35°C, and fluorescence was measured in 200-l aliquots of supernatant with a spectrophotometer at an excitation wavelength of 527 nm and an emission wavelength of 555 nm (Synergy HT; BioTek, Winooski, VT). ABC-specific rhodamine efflux was calculated for each strain by subtracting the fluorescence in glucose-free supernatants from that in glucose-containing supernatants. To enhance interexperimental reproducibility, the rhodamine efflux of each strain was normalized to that of C. glabrata CBS15126, which was tested in each experiment. In separate experiments, the cellular accumulation of rhodamine 6G in C. glabrata strains was determined by fluorescence microscopy. Rhodamine 6G uptake and efflux were achieved as described above. After a 15-min incubation, suspensions were centrifuged and pellets were washed twice in PBS and observed under a triple-band fluorescence microscope (Olympus BX-51; Olympus, Melville, NY) with the rhodamine filter. Images were acquired with an Olympus DP71 camera (Olympus, Tokyo, Japan) and Cell^A software with the exposure kept constant at 74 ms.
ABC transporter gene expression. We assessed the association of the PAP phenotype with the expression of efflux transporter genes known to be associated with azole resistance in C. glabrata by real-time reverse transcription-quantitative PCR (RT-qPCR) as previously described (5). Isolates were grown in liquid YAG at 35°C in a shaking incubator to mid-exponential phase. Yeast cells were enumerated in a hemocytometer and transferred into fresh YAG with fluconazole at 8 or 64 mg/liter or into drug-free YAG at a final concentration of 10 7 /ml. Yeast suspensions were incubated for 3.5 h at 35°C with shaking; this stage did not affect C. glabrata viability, as determined by quantitative subculture on Sabouraud's agar. Next, cells were pelleted, snap-frozen in liquid nitrogen, and thoroughly ground with 3-mm glass beads. Lysis buffer (Qiagen, Hilden, Germany) was added, and the lysate was further homogenized in a TissueLyser (Qiagen). RNA was extracted with the RNeasy kit (Qiagen) according to the manufacturer's instructions, including in-column DNase treatment. cDNA was generated from 50 to 100 ng of total RNA with the High Capacity RNA-to-cDNA kit (Applied Biosystems/Life Technologies, Grand Island, NY). cDNA was diluted 1:10 in RNase/ DNase-free water and subjected to RT-qPCR with primers and TaqMan probes for CDR1, SNQ2, PDH1, and PDR1 (see Table S1 in the supplemental material). Each 25-l reaction mixture included 5 l of a cDNA template, 12.5 l of ABI universal master mix, 0.4 M each primer, and 0.2 M TaqMan probe. PCRs were run in an ABI Prism 7500 sequence detection system (Applied Biosystems/Life Technologies) under the following conditions: 95°C for 10 min (AmpliTaq gold activation), followed by 40 cycles of 95°C for 15 s (denaturation) and 60°C for 60 s (annealing and extension). The amplification efficiency was determined for each gene. For each target gene, the mean threshold cycle (C T ) was calculated from triplicate experiments and normalized to ACT1 expression (⌬C T ). Each gene was calibrated to the ACT1-normalized expression of its counterpart in nonheteroresistant C. glabrata strains CBS 15126 and Cg1775 (⌬⌬C T ), and fold expression was calculated as 2 Ϫ⌬⌬CT . A 2-fold increase in expression was arbitrarily set as the threshold for significant upregulation (5). The PDR1 gene sequences of six FLC HR and four FLC N C. glabrata strains were analyzed (see Text S1 in the supplemental material).
Persistent infection in a mouse model of hematogenous disseminated candidiasis. C. glabrata is nonlethal when injected intravenously into immunocompetent mice; nonetheless, it can persist for prolonged periods in visceral tissue (34). To determine if the PAP phenotype is associated with persistent infection during fluconazole treatment in vivo, we compared the tissue burdens of Cg1775 (FLC N , fluconazole MIC of 0.25 mg/liter) and Cg1646 (FLC HR , fluconazole MIC of 1 mg/liter) in the kidneys of fluconazole-treated and untreated BALB/c mice. Cg1708 was used as a reference fluconazole-resistant strain. Eight-week-old female BALB/c mice weighing 20 to 25 g (Harlan, Rehovot, Israel) were infected by tail vein injection of 4 ϫ 10 7 cells in groups of 10 per C. glabrata strain. Five mice per strain were treated with fluconazole (Pfizer, New York) at 100 mg/kg once a day intraperitoneally. The remaining five mice were injected daily with a similar volume of sterile saline. Treatment was continued for 7 days postinfection, after which mice were sacrificed by CO 2 inhalation and their kidneys were harvested and processed for quantitative determination of fungal burdens. Organs were weighed and homogenized in 1 ml of sterile saline with a TissueLyser set to 1 min at 50 Hz. Homogenates were serially diluted 10-fold and inoculated onto YAG plates for 48 h at 30°C. Fungal burdens were calculated and expressed as numbers of CFU per gram of tissue. Experiments were repeated three times.
Phylogenetic analysis. Phylogenetic relationships between different C. glabrata strains were resolved by sequencing and analysis of the rapidly Persistence of viable C. glabrata in homogenized kidneys of immunocompetent BALB/c mice was determined 7 days after tail vein infection. Diamonds, Cg1775 (FLC N isolate); triangles, Cg1646 (FLC HR isolate); squares, Cg1708 (FLC R isolate); empty symbols, mice injected intraperitoneally with fluconazole 100 mg/kg/day; full symbols, mice treated with saline control. Values at the top are the numbers of mice with persistence of viable C. glabrata in the kidneys at the end of 7 days of fluconazole treatment, the total number of mice in each group, and the percentage with persistent infection. evolving IGS region between nuclear genes CDH1 and ERP6 on chromosome A as described previously (35). Phylogenetic analysis included 48 fluconazole-susceptible C. glabrata isolates, 6 fluconazole-resistant clinical isolates, and reference C. glabrata strains CBS138, CBS15126, and BG2. After genomic DNA extraction, an~660-bp fragment of the IGS was amplified with primers 00605 (5= C TCA CAA ATG GAT TCC TTA AAG AGT TCG 3=) and 00627 (5= GT C ACC AGA GTT GGA GTA CAT GTA G 3=). PCR was performed with 50-l tubes containing 400 nM each primer under the following conditions: initial denaturation at 95°C (4.5 min) and then 35 cycles of 95°C (45 s), 52°C (1 min), and 72°C (1 min), followed by 72°C (7 min). The PCR products were sequenced with primers 00605 and 00627. Sequences were aligned with Clustal Omega (36), and phylogenetic trees were constructed in MrBayes 3.2 (37) by Bayesian Markov chain Monte Carlo (MCMC) methodology. The average standard deviation of split frequencies (ASDSF) was calculated by comparing split and clade frequencies across 1 ϫ 10 6 MCMC runs started from different randomly chosen trees. The ASDSF cutoff was set at Ͻ0.05. The final output is presented as a cladogram with the posterior probabilities for all of the splits.
Statistics. AUCRs of FLC HR and FLC N strains were compared with the Mann-Whitney test. Comparison of transporter gene transcript and rhodamine 6G fluorescence values was done by one-sided analysis of variance with the Bonferroni post hoc test to compare specific pairs of values. Correlations among HR, transporter gene expression, and rhodamine 6G efflux were explored by using linear regression. Fungal persistence in mice infected with C. glabrata was observed as a dichotomous outcome. Therefore, we compared groups of mice infected with different C. glabrata strains by considering fungal clearance a categorical variable. A two-sided type I error of 0.05 was used as the threshold for statistical significance. Calculations were performed in Prism 6.0 (GraphPad, San Diego, CA) and Stata 11.2 (StataCorp, College Station, TX).
Data availability. Representative C. glabrata strains (10 FLC HR and 9 FLC N ) have been deposited at the CBS-KNAW Fungal Biodiversity Centre, Utrecht, Netherlands (http://www.cbs.knaw.nl/; see Table S1 in the supplemental material). The GenBank accession numbers of the CDH1-ERP6 IGS sequences used for phylogenetic analyses are listed in Table S1. The GenBank accession numbers of the PDR1 gene sequences are listed in Table S2.