Clonal Population of Flucytosine-Resistant Candida tropicalis from Blood Cultures, Paris, France

Such isolates are widespread around clinical centers and are associated with malignancies that cause fewer deaths than other C. tropicalis isolates.

isolates were compared on the basis of several phenotypic and molecular features.

Strains
Clinical isolates of C. tropicalis recovered from blood cultures during the YEASTS program from October 1, 2002, through September 30, 2006, were selected for the study. Epidemiologic and clinical data concerning the patients were collected by using a standardized electronic form. Isolates (1 isolate/patient) were sent to NRCMA for identifi cation and MIC determination (see below). All isolates were stored frozen in 40% glycerol at -80°C.
The type strain of C. tropicalis CBS 94 (ATCC 750, S 5FC) was included in the study as a reference. In addition, 29 strains of taxonomic synonyms available at the Centraalbureau voor Schimmelcultures (CBS, Utrecht, the Netherlands) were studied.

Phenotypic Characterization of All C. tropicalis Isolates
All isolates were identifi ed at the species level by using the assimilation patterns obtained with the commercialized strips ID32C (bioMérieux, Marcy-l'Etoile, France). MICs to 9 systemic antifungal agents were determined for all clinical isolates and the type strain by using the EU-CAST microdilution method (4). For nonclinical isolates, only MICs of 5FC and fl uconazole were determined.

Growth Characteristics
For the fi rst 16 S 5FC and 14 R 5FC consecutive isolates of C. tropicalis and for CBS 94 other studies were performed. Additional carbon sources were tested by using the commercial strips CH50 (bioMérieux). Maximal temperature of growth (42°C or 45°C) was determined on Sabouraud dextrose agar. Growth in hyperosmolar medium (50% glucose or 10% NaCl) was also evaluated.

Nucleotide Sequence Determination
After 24 hours of incubation at 27°C on Sabouraud agar plates, single colonies were discharged in 1 mL of distilled water in a microcentrifuge tube, and DNA extraction was performed by using the High Pure PCR Template Preparation Kit (Roche Applied Science, Mannheim, Germany) according to manufacturer's instructions. Universal fungal primers were used for the amplifi cation of the internal transcribed spacer 1 (ITS1)-5.8S-ITS2 (primers V9D and LS266 [5,6]) and the 26S (primers NL1 and NL4 [7]) rDNA regions. Primers (Table 1) were designed to amplify a partial sequence of the actin gene (GenBank accession nos. AJ389059 and AJ508499) and the 14-α-demethylase gene (GenBank accession nos. AY942646 and M23673). Reaction volumes of 20 μL contained 1 μL of genomic DNA, 1.25 U of AmpliTaq Gold, 2 L of PCR buffer 10×, 2 μL of 25 mmol/L MgCl 2 2 μL of 2.5 mmol/L deoxyribonucleoside triphosphates (dNTPs) (Roche), and 1 μL ac.uk/clustalw/). Following preliminary results, primers were then designed to amplify the complete sequence of the orotidine-5′-phosphate decarboxylase gene (URA3, GenBank accession no. AF040702) ( Table 1). Furthermore, the complete sequences of FCY1 (coding for the cytosine deaminase), FCY2 (coding for the purine cytosine permease), and FUR1 (coding for the uracil phosphoribosyl transferase) were determined. Primers were designed by using sequences from the Broad Institute C. tropicalis database genome (locus CTRG_02927.3 for FCY1, locus CTRG_02059.3 for FCY2, and locus CTRG_02689.3 for FUR1) ( Table 1). The sequences were amplifi ed as described above (except for the duration of annealing and elongation [1 min] when using the primers set CTCP1f/CTCP1r). The sequences were translated with the standard genetic code (www.bioinformatics.org/sms/index.html). The resulting protein sequences were aligned with the BioloMICS software (BioloMICS, version 7.2.5, BioAware S.A., Hannut, Belgium).

Microsatellite Selection
C. tropicalis genome sequences available from Gen-Bank databases and from the Broad Institute (www.broad. mit.edu/annotation/fungi/candida_tropicalis) were studied to identify sequences containing microsatellite repeats. Two polymorphic microsatellite markers (PMMs) were selected, 1 upstream of the URA3 gene (URA3 PMM) and 1 on a nonannotated sequence (CT14 PMM). Oligonucleotide primers were designed from the sequence of the corresponding fl anking regions to obtain PCR products ranging in size from 100 bp to 200 bp. One primer of each set was 5′ labeled with different dyes (Table 1). PCR was conducted independently for the 2 loci in a 20-μL reaction volume containing 2 μL of extracted DNA, 1.25 U of Am-pliTaq Gold, 2 μL of PCR Buffer 10×, 4 μL of 25 mmol/L MgCl 2 , 2 μL of 2 mmol/L dNTPs, and 0.2 μL (10 μM) of primers. PCR amplifi cations were performed for a total of 27 cycles by using the following conditions: denaturation at 95°C for 30 s, annealing at 55°C for 30 s, extension at 72°C for 1 min, and a fi nal extension step of 5 min at 72°C. Two microliters of each PCR product mixed with 20 μL of formamide and 0.5 μL of an internal standard labeled with 6-carboxy-X-rhodamine dye (GeneScan-500 Tamra, Applied Biosystems) was run on an ABI Prism 310 Genetic Analyzer (Applied Biosystems). Sizes of the allele and PCR fragments were determined with GenScan 3.0 (Applied Biosystems, Weiterstadt, Germany). To assign a specifi c length to a PCR fragment, all electromorphs were aligned with that of the type strain (CBS 94). Each allele was named after the length of PCR fragments. Isolates for which 1 signal was observed for a given locus on the electromorph were considered homozygous for this locus by analogy with what is reported for another diploid yeast, C. albicans (8).

Multilocus Sequence Typing Analysis
Three of the 6 MLST loci recently described were analyzed as reported in Tavanti et al. (9). These loci were selected because, according to these authors, they were associated with more polymorphism (XYR1 and SAPT4) and with antifungal resistance (MDR1). Both strands of purifi ed amplifi ed fragments were sequenced, and sequences were edited as described above. Heterozygosity was defi ned by the presence of 2 coincident peaks of similar height in the forward and the reverse sequence chromatograms. The 1-letter code for nucleotides from the nomenclature of the International Union of Pure and Applied Chemistry (IUPAC, www.bioinformatics.org/sms/iupac.html) was used. Sequences were compared with the allele sequences of the C. tropicalis MLST database (www.pubmlst.org/ ctropicalis). For each gene, distinct alleles were identifi ed and numbered by using the Internet-based MLST program (www.mlst.net). New alleles were submitted to the MLST C. tropicalis database.

Statistical Analysis
In accordance with French regulations, the clinical database was approved by the Commission Nationale de l'Informatique et des Libertés. Information concerning demographic data, risk factors for candidiasis, and outcome 30 days after the diagnosis of fungemia were recorded. We considered 3 groups of patients according to the infecting isolate: S 5FC, R 5FC that belong to the clone ( R 5FC clone, see below), and R 5FC that do not belong to the clone (termed "other R 5FC"). The sociodemographic and clinical characteristics were compared between the 3 groups of isolates by using the Fisher exact test. The χ 2 Armitage trend test (10) was used to assess a trend in the evolution of the R 5FC clone's proportion among resistant strains across years of study. Multinomial logistic regression (11) adjusted on clinical center was used to investigate the factors associated with infection by the R 5FC clone or other R 5FC isolates compared to S 5FC isolates according to sociodemographic and clinical characteristics. A logistic regression model adjusted on clinical center was also performed to identify the factors associated with the acquisition of the R 5FC clone compared with other R 5FC isolates. Regression models were constructed by using the backward procedure. First, all covariates with a p value <0.25 in univariate models were simultaneously entered into the regression model. The set of covariates with the largest p value was iteratively removed from the model until all of the covariates (or blocks of covariates) remaining in the reduced model had a p value <0.05. Statistical analyses were performed with Stata software, version 9.0 (StataCorp, College Station, TX, USA).

Phenotypic Characterization of R 5FC Isolates
We analyzed the episodes of fungemia caused by C. tropicalis and recorded during the fi rst 4 years of the YEASTS study; 130 episodes were recorded in 24 of the 27 participating centers. Distribution of fl ucytosine MICs showed 2 populations, 1 with MICs <2 μg/mL and 1 with MICs >8 μg/mL (Figure). In light of these results, susceptibility to 5FC ( S 5FC) was defi ned by an MIC <8 μg/mL and resistance ( R 5FC) by an MIC >8 μg/mL.
The proportion of R 5FC isolates (45 [35%]) of the 130 isolates) was uneven, ranging from 0% to 67% of the isolates, depending on the center of isolation. However, the proportion of R 5FC isolates did not differ over the study period (data not shown). We fi rst studied the characteristics of a subset of 16 S 5FC and 14 R 5FC C. tropicalis isolates (strain CBS 94 had all the characteristics of S 5FC clinical isolates described below but is not included in the analysis). There was no difference in terms of growth in hyperosmolar media between R 5FC and S 5FC clinical isolates. By contrast, R 5FC and S 5FC isolates differed in the proportion of isolates growing at 45°C (40% vs. 100%, p<0.001) and assimilating starch (12.5% vs. 50%, p = 0.054) and xylitol (62.5% vs. 12.5%, p = 0.009). No difference in the MIC of azoles or caspofungin was noted. All 29 strains of C. tropicalis synonyms stored at the CBS exhibited 5FC MICs <0.5 μg/mL.

Genotypic Characterization of R 5FC Isolates
This subset of isolates (16 S 5FC and 14 R 5FC) was further analyzed. The deletion of 1 nucleotide (A) in position 106 (according to the type strain sequence, Gen-Bank accession no. AY939810) of the ITS2 region was observed in 14 (100%) of the 14 R 5FC isolates compared to 5 (32%) of the 16 S 5FC isolates (p<0.001) (GenBank accession no. EU288196). No difference in nucleotide sequences was found for the D1/D2 region of the 26S rDNA or in the portion of the 14-α demethylase (490 bp) and actin (550 bp) genes analyzed. PMM results are summarized in Table 2. The URA3 and CT14 PMM led to 6 and 7 different allelic associations, respectively. The association of both markers led to 13 PMM profi les. The 14 R 5FC isolates had the same URA3/CT14 PMM profi le; however, among the 16 S 5FC, 15 had a PMM profi le different from that of the R 5FC, and 1 (ODL6-560) had the R 5FC PMM profi le. All but 1 (ODL2-237) of the R 5FC isolates had the same MLST profi le, whereas none of the S 5FC isolates exhibited the same combination of the 3 MLST studied. The entire URA3 nucleotidic sequence of the translated region was identical except in 1 position (GenBank accession no. EU288194 for the type strain CBS 94 and EU288195 for 1 of the R 5FC isolates). S 5FC isolates were either homozygous or heterozygous at position 529 (A-A or A-G), whereas all the R 5FC isolates were homozygous G-G. This produced a change of K177E (lysine → glutamate) for the R 5FC isolates.
Thus, results obtained with nucleotide changes (deletion of A in position 106 in the ITS2 region, mutation in position 529 of the URA3 gene) and polymorphisms in 3 MLST and 2 PMMs suggested that the 14 R 5FC studied were clonal. We thus decided to use the 2 PMMs (URA3 PMM, CT14 PMM) to genotype all the C. tropicalis isolates recovered during the study period in the YEASTS program. Of the 130 C. tropicalis isolates (including the 30 isolates studied above), 45 were R 5FC (Table 3). Thirtythree different profi les were observed when both PMMs were combined for the 130 isolates. Among the 45 R 5FC, a total of 29 isolates exhibited the profi le associated with the R 5FC clone; 16 were different, with 11 different profi les, 6 of which were shared with S 5FC isolates. Among the S 5FC isolates, 4 had the PMM profi le associated with the R 5FC clone. The URA3 gene was sequenced for these 4 isolates, and none had a G in position 529.
We then studied genes potentially involved in the mechanisms of 5FC resistance. For the FUR1 sequences, no missense mutation was observed, and the complete coding sequence of the type strain CBS 94 (GenBank accession no. EU327978) was similar to the sequence of the R 5FC clone; however, a few silent mutations were observed in a few S 5FC isolates (GenBank accession nos. EU327979, EU327980, and EU327981). Concerning the cytosine deaminase sequences (FCY1), only 1 silent mutation, C21T, occurred for the R 5FC clone (GenBank accession no. EU327982). Finally, for the purine cytosine permease (FCY2), the sequences of the type strain (GenBank accession no. EU327983) and of the R 5FC clone were similar. A few heterozygosities were observed for the S 5FC isolates (GenBank accession nos. EU327984 and EU327985), but all these mutations were silent.

Factors Associated with Fungemia Caused by the C. tropicalis R 5FC Clone
All 130 isolates corresponded to incident fungemia in different persons. The R 5FC clone was recovered during the 4 years of study with a trend toward a decreased proportion over time ( [44%]) during the fi rst, second, third, and fourth year of the study, respectively; p = 0.06). The proportion of the clone also varied across clinical centers (data not shown). Factors associated with fungemia caused by the R 5FC clone of C. tropicalis were analyzed ( Table 4). The proportion of patients infected by the R 5FC clone was signifi cantly higher among patients with malignancies but their death rate was signifi cantly lower than for patients infection with other R 5FC or S 5FC isolates. Multinomial logistic regression was adjusted by clinical centers to investigate the factors associated with infection by the R 5FC clone or others R 5FC isolates compared with S 5FC isolates. The risk of being infected by the R 5FC clone compared with a S 5FC isolate signifi cantly increased in case of malignancy (odds

Discussion
The bimodal distribution of 5FC MICs against C. tropicalis isolates prospectively collected from 27 different clinical centers in the Paris area (YEASTS program, Figure) suggested that the C. tropicalis population was heterogenous. On the basis of physiologic characteristics and molecular analysis (nucleotide sequences of the ITS regions, D1/D2 region of the large subunit, and large portions of the actin and 14-α-demethylase genes showing >99% similarity), we fi rst assessed a subset of isolates (the fi rst consecutive 14 R 5FC and 16 S 5FC isolates) and determined that both populations belong to the same species.
All 14 R 5FC isolates had a single nucleotide deletion in position 106 of the ITS2 region, although 5 of the 16 S 5FC isolates harbored it. When additional genotypic markers were used, all 14 R 5FC isolates had the same allelic combination for 2 PMMs selected (URA3 and CT14), the same missense mutation in the URA3 gene, and the same diploid sequences for the 3 MLST loci studied. By contrast, only 1 of 16 S 5FC isolates had the same PMM profi les as the R 5FC isolates, but this isolate differed in its MLST profi le and the lack of mutation in the URA3 gene. In addition, the 16 S 5FC isolates exhibited 16 different MLST and 12 different PMM profi les. This fi nding suggested the existence of a R 5FC clone, but the rest of the population was genetically diverse. We thus analyzed the 130 isolates of C. tropicalis collected over 4 years in the YEASTS program by using the 2 PMMs and sequenced the URA3 gene when the PMM  profi le was identical to that of the 14 R 5FC isolates previously studied. We discovered that 29 (64%) of 45 R 5FC isolates had an identical PMM profi le ( R 5FC clone), while 11 and 30 different PMM profi les were found among the other 16 R 5FC isolates and the 85 S 5FC isolates, respectively. According to these data, we assumed that these 29 R 5FC isolates were clonal or at least highly genetically related. The proportion of 35% of C. tropicalis isolates resistant to 5FC is unusual (3). Other studies report between 0 (12) and 15% (13) with intermediate values (14)(15)(16), and all the isolates stored as C. tropicalis in the CBS collection since 1912 exhibited 5FC MIC <0.5 μg/mL. When we started the YEASTS program in 2002, the proportion of R 5FC was already at 46% and the clone accounted for 85% of the R 5FC isolates. The trend test suggested that the dispersal of the clone is declining in the Paris area. Whether this decline is specifi c for blood isolates or is a geographically and temporally restricted phenomenon deserves evaluation by using isolates collected over time from various body sites and geographic areas. An old report on isolates collected from various regions of France established with a nonstandardized technique that as many as 70% of the 63 isolates tested had an MIC >32 μg/mL in the 1980s (17), and a recent study from Germany on clinical isolates recovered from various body sites including blood reported that 58.3% of isolates were resistant (18). Whether any of these isolates belong to the R 5FC clone would be of interest. Of note, a recent study of 104 C. tropicalis clinical isolates recovered from various countries (9) showed that none of the 5 R 5FC isolates collected in the United Kingdom has the MLST profi le of the R 5FC clone.
In the univariate analysis, patients infected by the S 5FC or R 5FC isolates or by the R 5FC clone differed signifi cantly in terms of proportion of underlying malignancies (higher in patients with the clone) and death rate 30 days after fungemia (lower for patients infected by the clone). The R 5FC isolates as a whole, and the R 5FC clone specifi cally, were unevenly distributed around the Paris area. When adjusted for clinical center, logistic regression analysis showed that, compared to infection by S 5FC isolates, no factor was independently associated with infection by R 5FC isolates other than the clone, whereas 2 parameters were associated with infection by the clone. Indeed, malignancies multiplied the risk of being infected by the clone by almost 4 and the risk for death was divided by 3 in case of infection with the clone. C. tropicalis fungemia, independent of susceptibility to fl ucytosine, has already been associated with hematologic malignancies (19,20) (unpub. data from the YEASTS group). Whether the R 5FC clone is less virulent, as established for C. albicans isolates with decreased susceptibility to 5FC, remains to be determined (21).
The resistance to 5FC was associated with the K177E mutation in the URA3 gene in the clone. The mechanism of 5FC action is a consequence of intrafungal formation of 2 metabolites, 5-fl uorodeoxyuridine monophosphate and 5-fl uorouridine triphosphate, which alter DNA and protein synthesis (22). The URA3 enzyme (orotidine 5′-phosphate decarboxylase, ODCase) is involved in the metabolic pathway of uridyl-monophosphate (UMP), which is a substrate of thymidylate synthetase and UMP kinase, both involved in nucleic acid synthesis. This mutation involves an amino acid already known to be variable among reference strains (e.g., ATCC 20336), but it has not been associated with modifi cation of the URA3 properties thus far (T. Noël, pers. comm.). Nevertheless, this mutation could, for example, modify the tridimensional structure of the protein, thereby affecting the binding affi nity of the substrate for the catalytic site and thus modifying the ODCase effi cacy. The ODCase is not known to interfere directly with 5FC activity. However, one of the resistance mechanisms against 5FC consists in increasing the transcription of all the genes involved in the de novo pyrimidine biosynthetic pathway (including URA3) to overproduce UMP (23).
The K177E mutation is associated with a specifi c PMM upstream the gene, with a possible role in the level of transcription. The fact that this PMM is homozygous may be due to a loss of heterozygosity. This phenomenon has been recently reported for C. albicans and the resistance to azoles (24) and for a specifi c C. albicans isolate and the resistance to caspofungin (25). The mutations described in the 5FC resistance of C. albicans (26) or C. lusitaniae (27) involved 3 major genes: FCY2 coding for the purine cytosine permease, which enables 5FC to enter the fungal cell; FCY1 coding for the cytosine deaminase, which transforms 5FC into 5FU; and FUR1 coding for the uracil phosphoribosyl transferase, which transforms 5FU into 5FUMP. The fact that the clone was susceptible to 5FU suggests that the 5FC resistance could result from a mutation in the cytosine deaminase, the cytosine permease, or both (23). However, the sequences of the R 5FC clone, some S 5FC isolates, and the type strain CBS94 did not show any mutation in coding sequences of FCY1, FCY2, or FUR1 susceptible to explain the resistance of the clone to 5FC. The mechanism explaining the possible relationship between the specifi c PMM (URA3 178/178 and CT14 148/151), the K177E mutation, and the resistance to 5FC remains to be determined.
Our results suggest that a clone of R 5FC isolates responsible for fungemia is widespread among patients hospitalized with malignancies in the Paris area and is associated with a lower mortality than that of other C. tropicalis isolates. Despite a trend toward a decreased proportion over time, further studies are needed to assess this clone's geographic and temporal distribution. Analysis of the 2 PMMs described in this study, coupled with determination of nucleotide at position 529 in the URA3 gene, should provide reliable means to track this clone.
Ms Desnos-Ollivier is an engineer at the National Reference Center of Mycoses and Antifungal at the Pasteur Institute in Paris, France. Her research interests include studying genotyping, sensibility to antifungal agents, and physiology of the yeasts responsible for human fongemia and underlining the relationship between genomic and phenotypic data.