Recent Increase in the Prevalence of Fluconazole-Non-susceptible Candida tropicalis Blood Isolates in Turkey: Clinical Implication of Azole-Non-susceptible and Fluconazole Tolerant Phenotypes and Genotyping

Candida tropicalis is the fourth leading cause of candidemia in Turkey. Although C. tropicalis isolates from 1997 to 2017 were characterized as fully susceptible to antifungals, the increasing global prevalence of azole-non-susceptible (ANS) C. tropicalis and the association between high fluconazole tolerance (HFT) and fluconazole therapeutic failure (FTF) prompted us to re-evaluate azole susceptibility of C. tropicalis in Turkey. In this study, 161 C. tropicalis blood isolates from seven clinical centers were identified by ITS rDNA sequencing, genotyped by multilocus microsatellite typing, and tested for susceptibility to five azoles, two echinocandins, and amphotericin B (AMB); antifungal resistance mechanisms were assessed by sequencing of ERG11 and FKS1 genes. The results indicated that C. tropicalis isolates, which belonged to 125 genotypes grouped into 11 clusters, were fully susceptible to echinocandins and AMB; however, 18.6% of them had the ANS phenotype but only two carried the ANS-conferring mutation (Y132F). HFT was recorded in 52 isolates, 10 of which were also ANS. Large proportions of patients infected with ANS and HFT isolates (89 and 40.7%, respectively) showed FTF. Patients infected with azole-susceptible or ANS isolates did not differ in mortality, which, however, was significantly lower for those infected with HFT isolates (P = 0.007). There were significant differences in mortality (P = 0.02), ANS (P = 0.012), and HFT (P = 0.007) among genotype clusters. The alarming increase in the prevalence of C. tropicalis blood isolates with ANS and HFT in Turkey and the notable FTF rate should be a matter of public health concern.


INTRODUCTION
Candida tropicalis is the leading cause of candidemia in India, Algeria, and Tunisia (Sellami et al., 2011;Chakrabarti et al., 2014;Megri et al., 2020) and the second to third leading cause of candidemia in South Asian countries Liu et al., 2019). Candida tropicalis is associated with a high mortality rate and the poorest prognosis among non-albicans Candida (NAC) species (Ko et al., 2019). Candida tropicalis has been considered an azole-susceptible NAC species; however, recent studies performed in several countries and worldwide indicate the emergence of azole-resistant C. tropicalis isolates in clinical settings (Arendrup et al., 2011;Chakrabarti et al., 2014;Choi et al., 2016;Fan et al., 2018;Xiao et al., 2018;Chen et al., 2019;Pfaller et al., 2019;Megri et al., 2020). According to a SENTRY study, C. tropicalis, similar to C. glabrata, also demonstrates an increased tendency to echinocandin resistance compared to other Candida species, although its rate is lower than that to azoles (Pfaller et al., 2019). Furthermore, a national candidemia study conducted in India indicates the emergence of multidrugresistant (MDR) C. tropicalis isolates in a similar proportion as what observed for MDR Candida auris (Chakrabarti et al., 2014). An additional matter of concern is the development of drug tolerance, which may allow the fungus to acquire stable genetic alterations leading to antifungal resistance and which could be potentially linked to azole therapeutic failure and mortality (Rosenberg et al., 2018;Berman and Krysan, 2020). Indeed, several studies indicate that C. tropicalis is among the most azoletolerant Candida species (Arthington-Skaggs et al., 2002;Rueda et al., 2017) and that fluconazole efficacy against C. tropicalis blood isolates with high azole tolerance is decreased, which can cause fluconazole therapeutic failure (FTF) (Astvad et al., 2018). Considering the limited number of antifungals available to treat candidemia, development of resistance against a single or multiple antifungals should pose a serious threat to patients with candidemia due to C. tropicalis, especially in developing countries where fluconazole is the most widely used antifungals (Singh et al., 2018;Arastehfar et al., 2019a,b).
Although our understanding of azole resistance mechanisms in C. tropicalis is limited, molecular studies indicate that mutations in the ergosterol biosynthesis gene ERG11, which result in reduced affinity of Erg11 to azoles, and/or overexpression of Erg11 and efflux pumps Cdr1 and Mdr1 could be the major contributors to azole resistance (Arastehfar et al., 2020g). At the same time, azole tolerance may depend on changes in the components of stress-response pathways (Rosenberg et al., 2018;Berman and Krysan, 2020). Resistance to echinocandins is mainly caused by mutations in hotspots (HSs) 1 and 2 of the FKS1 gene encoding the catalytic subunit of β-(1,3)-glucan synthase (Arastehfar et al., 2020g). Genotyping techniques made it possible to determine the link between the genotype and fluconazole resistance and to reveal a propensity of fluconazole-resistant blood isolates of C. tropicalis to expand in clusters with a similar genetic background (Chen et al., 2019). Given the association between the genotype and mortality in other Candida species (Byun et al., 2018;Arastehfar et al., 2019a), it can be hypothesized that there are similar correlations between azole tolerance and resistance, patient mortality, and the genotype of C. tropicalis.
A recent study explored antifungal susceptibility patterns of Candida bloodstream isolates in Turkey  and reported the lack of drug resistance for C. tropicalis (Arikan-Akdagli et al., 2019). However, our single-and multi-center investigations conducted later in Turkey have revealed a very recent (from 2017 onward) emergence of azole/echinocandinresistant and even MDR Candida species (Arastehfar et al., 2020b,c,e), suggesting the existence of the same trend for C. tropicalis. Therefore, we conducted a comprehensive multicenter study to update the rate of antifungal resistance among C. tropicalis blood isolates recovered in 2017-2019 in Turkey; some isolates recovered before 2017 but not analyzed in the previous study (Arikan-Akdagli et al., 2019) were also included. The drug resistance mechanisms were explored by sequencing ERG11 and HS1/2 of FKS1. We also examined the level of fluconazole tolerance and assessed its correlation with azole exposure, FTF, and patient mortality. Multilocus microsatellite typing (MMT) was performed to determine the genetic relatedness among C. tropicalis bloodstream isolates and associations between genotype clusters and fluconazole minimum inhibitory concentrations (MICs), fluconazole tolerance, and patient outcome.

Multilocus Microsatellite Typing
Genotypic diversity of C. tropicalis isolates was evaluated according to six markers and 12 loci by using a previously published protocol (Wu et al., 2014). Microsatellite data were organized as categorical values; data analysis was performed by using Bionumerics software v7.6 (Applied Math, Sint-Martens-Latem, Belgium) and dendrograms were constructed according to the unweighted-pair group method using average linkages. Distinct genotypes were considered if two strains differed in more than one locus.
Among the azole antifungals used to treat patients, fluconazole was the main drug applied for both prophylaxis and targeted therapy; therefore, tolerance assessment was performed only for fluconazole. To explore the level of fluconazole tolerance, the mean growth rate above the MIC (4-64 µg/ml) was recorded after 48-h incubation as described previously (Rosenberg et al., 2018;Berman and Krysan, 2020). The level of tolerance was categorized as low-moderate or high when the growth at concentrations above the MIC was <50 and ≥50% compared to the drug-free control, respectively. ERG11 and HS1-and HS2-FKS1 Sequencing ERG11 and HS1 and HS2 sequencing was performed for all isolates using primers and conditions described previously (Arastehfar et al., 2020a). Contigs were assembled by SeqMan Pro (DNASTAR, Madison, WI, United States) and the resulting sequences were aligned by using MEGA software v7.0 (Kumar et al., 2016) with C. tropicalis MYA-3404 genome (AAFN00000000.2) as a WT reference (Butler et al., 2009).

Statistical Analysis
Statistical analysis was performed with SPSS v24 (SPSS Inc., Chicago, IL, United States). Given the high number of genotypes (n = 125), we increased the statistical power by grouping isolates into 11 clusters according to genotype similarity. The association between clusters and fluconazole MICs was determined by using chi-square and Kruskal Wallis tests (since the fluconazole MIC data were non-parametric) and the magnitude of association was calculated according to the Eta correlation coefficient. Chi-square test was also used to assess statistical association between the outcome (mortality/survival) and the genotype cluster, and logistic regression analysis was further used at R square values over 0.07. Patients infected with multiple isolates of various genotypes were not considered for the association between the genotype cluster and mortality. Correlation between azole tolerance and outcome was assessed by chi-square test. P values ≤ 0.05 were considered statistically significant.
Frontiers in Microbiology | www.frontiersin.org statistical analysis section, tolerance and outcome). There was no significant difference in mortality between patients infected with ANS and azole-susceptible isolates (two-tailed chi-square test, P = 0.987) (Supplementary Material, statistical analysis section, azole resistance and outcome).

DISCUSSION
In this study, we revealed a considerable rate of infection with ANS C. tropicalis in Turkey (15.7%), which is in contrast with a previous multi-center study, which did not detect any such isolates (0%) (Arikan-Akdagli et al., 2019). We also observed that HFT and ANS isolates could be encountered in azole-naïve as well as in fluconazole-treated patients and that ANS and HFT phenotypes, although with varying degrees, were associated with FTF. Analysis of microsatellite cluster distribution revealed its significant association with fluconazole MICs and tolerance, and patient mortality, which highlights the clinical importance of genotyping data. Consistent with previous study in Turkey (Arikan-Akdagli et al., 2019), the antifungal susceptibility profile of C. tropicalis updated in this study indicated the absence of isolates with echinocandin resistance and/or non-WT susceptibility to AMB. In contrast to a previous study reporting the absence of azole-resistant C. tropicalis (Arikan-Akdagli et al., 2019), however, we identified 18.6% ANS C. tropicalis isolates, among which 93 and 58% were cross-resistant to three azoles or fluconazole + voriconazole, respectively. Importantly, we observed an alarming rate (40-50%) of ANS isolates in some hospitals in 2019, which is also documented in other studies (Arendrup et al., 2011;Chen et al., 2019;Pfaller et al., 2019); these statistics require particular attention, especially in the countries where azoles are the main antifungals used for candidemia therapy (Sellami et al., 2011;Singh et al., 2018;Arastehfar et al., 2019a. Notable variations in the rate of azole-resistant isolates (0-100%) observed among the analyzed hospitals could be attributed to differences in intervention strategies, including the use of azoles and infection control practices. To address the resistance mechanism, we sequenced the ERG11 gene and unexpectedly found a very low rate of the Y132F mutation considered responsible for azole resistance (Fan et al., 2018), which was identified in only two azole-resistant isolates. However, many other factors can contribute to azole resistance (Fan et al., 2018;Arastehfar et al., 2020g); thus, a recent study did not identify accountable ERG11 mutations in fluconazole-resistant C. tropicalis blood isolates but instead found specific polymorphisms in transcription factors regulating the expression of ERG11 and efflux pump-related genes, i.e., UPC2, MRR1, and TAC1 (Arastehfar et al., 2020a). Although we detected other ERG11 mutations, they were found exclusively in azolesusceptible isolates, which is consistent with a previous report (Oliveira et al., 2013).
To understand the association between azole use and the emergence of ANS C. tropicalis isolates, we analyzed the clinical history of respective patients (20/127; 15.7%) and found that 50% of them were azole-naïve, which is similar to the results obtained for patients infected with Candida parapsilosis (Tóth et al., 2019) and C. auris (Jeffery-Smith et al., 2018). Recent studies from Iran (Arastehfar et al., 2020a), Taiwan (Chen et al., 2019), and Japan (Chong et al., 2012) have reported the same phenomenon, which could be the driving force behind the replacement of C. albicans as the first leading cause of candidemia with C. tropicalis documented in Taiwan (Tulyaprawat et al., 2020). We also found an association between fluconazole resistance and genotypes, which may account for the dramatic increase in the number and persistence of ANS isolates in clinical settings since the previous candidemia study in Turkey (Arikan-Akdagli et al., 2019), hypothetically because azole-resistant isolates may possess a higher mutagenesis potential allowing them to thrive in conditions when fluconazole is heavily used (Healey et al., 2016). The remaining ANS isolates were recovered from nine patients, almost 89% of them (8/9) showed FTF. Regarding the two patients infected with the C. tropicalis isolates harboring Y132F, one was azole-naïve (#30) and the other one in keeping with the other Candida species carrying Y132F (Arastehfar et al., 2020g) showed failure to both fluconazole and voriconazole for which the isolate was resistant against both agents (#30).
Although we did not detect echinocandin resistance and all C. tropicalis isolates studied had WT for FKS1, considering the global increase in echinocandin-resistant C. tropicalis (Pfaller et al., 2019), the emergence of such isolates in Turkey could be only a matter of time, especially in view of a significant increase in ANS isolates over the past few years observed in this study.
Analysis of fluconazole tolerance, which is known to be considerable among C. tropicalis isolates, revealed a lower tolerance rate (34%) compared to a previous study (73.5%) (Rueda et al., 2017); however, 96.2% of our isolates were HFT. Similar to patients infected with ANS isolates, a notable proportion (44%) of the patients infected with fluconazole-tolerant isolates were azole-naïve. Among the patients infected with fluconazole-tolerant C. tropicalis and treated with fluconazole/azoles, approximately 31% showed FTF; furthermore, all patients (100%) infected with HFT-ANS isolates had FTF. Although pharmacodynamics and other host-related factors affecting therapeutic success should be considered, our findings, which are consistent with earlier studies (Astvad et al., 2018;Rosenberg et al., 2018), highlight the importance of fluconazole tolerance assessment in clinical settings. We propose that C. tropicalis isolates exhibiting HFT (≥50->75% growth at concentrations over MIC) may predict azole therapeutic failure. Indeed, the clinical implication of tolerance in the bacteriology field is more appreciated, where it has been associated with prolonged antibiotic treatment, therapeutic failure, and rapid development of resistance (Levin-Reisman et al., 2017;Windels et al., 2019;Berti and Hirsch, 2020;Liu et al., 2020). More importantly, antibiotic tolerance can threaten the clinical efficacy of combination antibiotic therapy used to minimize the emergence of resistant bacteria and to enhance the efficacy of treatment (Liu et al., 2020). Similarly, emerging recent studies in medical mycology documented FTF among isolates showing high level of tolerance to this drug (Astvad et al., 2018;Rosenberg et al., 2018). Altogether, these studies point to the importance of measurement of tolerance among clinically obtained bacterial and fungal isolates to guide clinicians in choosing an appropriate antifungal regimen. Of note, collecting multiple azole susceptible C. tropicalis isolates with undetected azole tolerance from the same patients might be attributable to host-related factors rendering them susceptible to acquire persistent candidemia (Arastehfar et al., 2020g,f).
In agreement with a previous study (Rueda et al., 2017), we found that patients carrying HFT isolates had a significantly lower mortality rate than those carrying isolates without detectable tolerance, which could be associated with unknown genomic mutations affecting the fitness cost (Rueda et al., 2017;Rosenberg et al., 2018). In contrast, we did not find significant difference in mortality between patients infected with azoleresistant and azole-susceptible isolates, which could be due to a negligible or no fitness cost for acquiring drug resistance. In vivo testing is needed to verify this hypothesis.
Although a previous study has not revealed correlation between the genotype and fluconazole tolerance, it could be attributed to a limited number of genotyped C. tropicalis isolates: each of 46 identified genotypes was represented with only a single isolate (Marcos-Zambrano et al., 2016), which could have prevented establishing a statistically significant link. In this study, we genotyped 161 C. tropicalis isolates and, to increase the statistical power of correlation analysis, grouped them to clusters containing related genotypes, which resulted in finding an association between genotype clusters and fluconazole tolerance. Interestingly, the clusters significantly differed in the number of HFT isolates, which could be explained by genetic, physiological, and metabolic variations among the strains (Rosenberg et al., 2018;Berman and Krysan, 2020). We also revealed correlation between the genotype cluster and patient mortality, which has been previously shown for C. glabrata (Byun et al., 2018;Arastehfar et al., 2019a). The mortality rate observed in this study is similar to those reported for C. tropicalis in Iran (Arastehfar et al., 2020a), Italy (Montagna et al., 2013), and the United States (Andes et al., 2016), and to that observed for C. glabrata in Turkey (Arastehfar et al., 2020e). Collectively, these findings emphasize the importance of using genotyping techniques in clinical settings, which may provide insightful observations with predictive prognostic values.
This study had some limitations, the main of which was the lack of detailed treatment data (dosage and duration) and clinical information due to the retrospective nature of the study. Also, we did not examine the expression of efflux pumps and ERG11 or sequence their master regulator genes MRR1, TAC1, and UPC2, which will be the scope of future research. Lastly, the mutation prevention concentration experiments, which quantitatively explores the azole tolerance is lacking in this study.
In summary, the current study revealed notable and rapid emergence of ANS isolates in Turkey, which necessitates continuous monitoring of the antifungal resistance rate on a multi-center scale. Furthermore, our results point on the clinical utility of azole susceptibility/tolerance profiling and genotyping, which can provide clues regarding azole therapeutic failure or success.

CONCLUSION
Candida tropicalis is the fourth leading cause of candidemia in Turkey and the blood isolates recovered during 1997-2017 were entirely susceptible to all antifungals tested. The studies conducted in other countries, however, have shown a dramatic increase in the number of ANS C. tropicalis blood isolates. Moreover, C. tropicalis proved to be the most tolerant Candida species to fluconazole and those with HFT showed FTF when tested in vivo. Therefore, the scope of our study was to reevaluate the burden of antifungal resistance of C. tropicalis blood isolates on a multicenter scale in Turkey and to evaluate the influence of ANS and HFT on fluconazole exposure, FTF, and mortality. Moreover, we sequenced the HS and HS2 of FKS1 and ERG11 to evaluate the echinocandins and azole resistance mechanisms. Indeed, as expected we observed a significant increase in the number of ANS isolates from 2017 onward when compared to the previous study conducted in Turkey. Interestingly, patients infected with both ANS and HFT C. tropicalis blood isolates showed FTF and some microsatellite clusters significantly had a higher number of ANS and HFT isolates. Similarly, we observed that some clusters were associated with a higher mortality rate. Collectively, our study implicated the growing concern of ANS C. tropicalis blood isolates in Turkey and showed that HFT isolates, similar to ANS isolates, can cause FTF. Moreover, we showed that the application of genotyping techniques may have prognostic values with regards to mortality and FTF in candidemia settings.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.

ETHICS STATEMENT
This study was approved by Ege University Hospital Ethical Committee and was granted with the following ethical approval number (20-2T/30). Consent forms were obtained from patients and we used numerical codes for C. tropicalis isolates to anonymize the patient's identity. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.