Molecular epidemiology of invasive Candida albicans at a tertiary hospital in northern Taiwan from 2003 to 2011

Candida albicans is a common cause of bloodstream fungal infections in hospitalized patients. To investigate its epidemiology, multilocus sequence typing (MLST) was performed on 285 C. albicans bloodstream isolates from patients in Chang Gung Memorial Hospital at Linkou (CGMHL), Taiwan from 2003 to 2011. Among these isolates, the three major diploid sequence types (DSTs) were 693, 659, and 443 with 19, 16, and 13 isolates, respectively. The 179 DSTs were classiﬁed into 16 clades by unweighted pair-group method using arithmetic averages (UPGMA). other isolates.


Introduction
Candida albicans is an organism of the normal gut flora in humans. It is also an opportunistic pathogen and is the fourth most common cause of nosocomial bloodstream infections with a mortality rate of 30-60% [1][2][3][4][5]. The risk factors of candidemia include invasive surgeries such as dialysis [6], implantation of central venous catheter [7], diabetes [6], burns [8], human immunodeficiency virus (HIV) infections, immunosuppression due to chemotherapy [9], and use of steroid drugs or broad-spectrum antibiotics [7].
The increasing frequency of invasive candidiasis and its serious outcome demand more epidemiological studies. A well-accepted method for typing C. albicans isolates is the multilocus sequence typing (MLST) [10][11][12]. It is based on nucleotide sequence variations within the 300-to 400bp internal regions of seven housekeeping genes, including AAT1a, ACC1, ADP1, MPIb, SYA1, VPS13, and ZWF1b. A sequence variation in each locus is assigned an allele number. Combination of allele numbers of these seven genes constitutes a unique diploid sequence type (DST) of a C. albicans isolate. Because MLST analysis relies only on nucleotide sequencing, the information about C. albicans isolates can be exchanged around the world through a global database (http://calbicans.mlst.net) [12,13]. At least 2400 DSTs have been recorded in the C. albicans MLST database.
UPGMA (unweighted pair-group method using arithmetic averages) is a method that can be used to determine the phylogenetic relationship among C. albicans isolates [11,14]. Using this method, Gong et al. has recently classified 1500 C. albicans isolates into 18 distinct clades [14]. Clade 1 isolates distribute globally, whereas isolates from the Pacific Rim cluster mostly in clades 14 and 17. Clade 1 isolates in general have a higher acid phosphatase activity and are less susceptible to 5-fluorocytosine [11,15] and more salt tolerant [16]. Another method called eBURST (electronic Based upon Related Sequence Types) allows determination of patterns of evolutionary descent by grouping isolates that differ at one or two of the seven MLST alleles into clonal clusters [17].
In this study, we determined the DSTs of 285 C. albicans isolates causing bloodstream infections in Chang Gung Memorial Hospital at Linkou (CGMHL) and investigated the epidemiology of the isolates using both UPGMA and eBURST. We also determined antifungal susceptibility of these isolates. Results showed that isolates of DST 693/clonal complex 20/clade 3 are more resistant to fluconazole than other isolates.

Candida Albicans Isolates
A total of 285 bloodstream infection isolates from CGMHL obtained between 2003 and 2011, including all 72 from pediatric patients and 213 randomly selected from 1098 archived adult ICU isolates, were investigated in this study (Supplementary Table). Each isolate was collected only once from a patient within the hospital admission. All isolates were identified by MALDI-TOF mass spectrometry and germ tube formation methods or CHROMagar Candida (BD).

Multilocus Sequence Typing
MLST of C. albicans isolates were performed as described by Bougnoux et al. [12,13] A portion of each of AAT1a, ACC1, ADP1, MPIb, SYA1, VPS13, and ZWF1b genes was amplified by polymerase chain reaction (PCR), and the resulting PCR products were sequenced. Each nucleotide sequence thus generated was compared to those in the C. albicans MLST database (http://calbicans.mlst.net/) to obtain an allele number. Any sequence that does not match with any of the preexisting sequences was given a new allele number. The combination of the seven allele numbers defined a unique DST of an isolate.

UPGMA Analysis
To determine phylogenetic relatedness, DSTs of the 285 CGMHL isolates and 996 isolates with known clades retrieved from the C. albicans MLST database were analyzed by UPGMA as described previously [11,14]. Briefly, the sequences of each housekeeping gene of each isolate were aligned to reveal polymorphic bases. The seven MLST polymorphic sequences from each isolate were then concatenated into a single sequence. Each base of the combined sequence was rewritten with two letters representing a homozygous or heterozygous diploid sequence. The genetic relatedness of the transformed sequences were analyzed by the software MEGA version 6 to generate a dendrogram [18].

eBURST Analysis
The relationships among the 285 CGMHL isolates and all 2448 isolates in the MLST database (date accessed 10.01.14) were determined by eBURST (http://eburst.mlst.net/). Based on the seven allele numbers of each isolate, eBURST placed related isolates into a clonal complex (CC) and predicted the ancestral DST of each CC by calculating the frequency of each DST genotype. The results of eBURST were displayed as the most parsimonious pattern of each descent of the ancestral DST type.

Antifungal Susceptibility Testing
A commercially available dried colorimetric microdilution panel (Sensititre YeastOne, TREK Diagnostic Systems) was used for susceptibility testing of C. albicans isolates to 5flucytosine, amphotericin B, anidulafungin, caspofungin, fluconazole, itraconazole, micafungin, posaconazole, and voriconazole. Briefly, C. albicans isolates (1.5-8 × 10 2 cfu) were seeded in YeastOne medium containing antifungal agents and incubated at 35 • C without CO 2 for 24 hours. The minimum inhibition concentration (MIC) of each antifungal agent was determined according to the guideline provided by the kit. The clinical breakpoints for sensitive, intermediate, and resistant isolates and epidemiological cutoff values for wild-type and non-wild-type isolates for the antifungal agents were referenced to those of Pfaller et al. [19] MIC 50 and MIC 90 values of C. albicans isolates against each antifungal agent were also calculated. Isolates with a fluconazole MICs > = 0.5 μg/ml were defined as having a lower fluconazole susceptibility, that is, more resistant to fluconazole.

Statistical analysis
The χ 2 and Fisher exact tests were performed to compare genotype distributions and antibiotic susceptibility. A Pdistance < .05 was considered significant. Prism 5.0 software (GraphPad, San Diego) was used for the analysis.

MLST clade distribution of CGMHL C. albicans isolates
To compare the genotypes of the CGMHL isolates to those of global isolates with previously reported MLST clades, UPGMA phylogenetic analyses were performed. The MLST clade distribution of all C. albicans isolates is shown in Figure 1 and  Table 1). Within the same clade, the isolates from this study were clustered together, especially those in clades 3, 4, and 11 ( Fig. 1).

Discussion
In the current study, the epidemiology of 285 bloodstream isolates of C. albicans was investigated. Based on 3.40 (34 genotypes divided by 10 variable bases) genotypes per polymorphism, respectively. These two genes were also found to have the best discriminating power by Bougnoux et al. [21] The 285 CGMHL and 996 MLST reference isolates were classified into 18 clades by UPGMA. The CGMHL isolates (open circles in Fig. 1) were clustered in all MLST clades except clades 2 and 13. Isolates in these two clades are mostly found in Europe and Africa, and clade 13 was previously recognized as Candida africana by phenotying [11,22]. Clades 1-4 and 11 have been evidenced the most consistent during rapid expansion of the database, and clades 3 and 11 have been shown to be very close to each other [22]. The uneven distribution of the CGMHL isolates in clades 3, 4, and 11 (Fig. 1) suggests a close phylogenetic association of CGMHL isolates within the same clade. The population of CGMHL isolates in clade 1 and singlets (isolates that could not be classified into any clade by UPGMA with a cutoff value of P = .04) was much smaller but that in clade 16 was significantly bigger than that of the global isolates published in 2007 [11]. The populations of CGMHL isolates in clades 3 and 4 were bigger than those of other Asian isolates, but not those of global isolates collected since 2000 (Table 1). In contrast, the number of CGMHL isolates in clade 17 was higher than that of global (both before and after 2000) but not of other Asian isolates.
UPGMA measures the P-distance of polymorphic nucleotide sequences. Although it provides a simple view of phylogenetic relationship of the isolates, some minor clades were altered when the isolate number increased [22]. Therefore, eBURST, another powerful algorism to reveal the genetic relationship of isolates, was also used in this study. Results showed that the percentages of CC8, CC20, CC15, and CC17 CGMHL isolates (15.8%, 9.5%, 8.4%, and 8.1%, respectively) were significantly higher than those of other Asian (5.8%, 1.7%, 0.6%, and 1.7%) and global isolates (1.6%, 0.4%, 0.3%, and 0.4%), suggesting an expansion of CC8, CC20, CC15, and CC17 isolates in CGMHL ( Table 2). Results of this study also showed that the CGMHL isolates in the same eBURST clonal complexes were grouped together in the same UPGMA clades (Table 3). Thus, there was a good correlation between UPGMA grouping and eBURST clustering. DST 659 was the predicted founder of the largest CC in the CGMHL isolates. DST 659 was determined to be CC11 (i.e., 11 th largest eBURST cluster) by Odds et al. [11], but was determined to be CC8 (i.e., 8 th largest eBURST cluster) in this study (Supplementary Table). In addition, DST 693, which was previously classified as a member of CC35 [11], was determined to be CC20 in this study (Supplementary Table). Both DST 659 and DST 693 clusters were greatly expanded during 2000-2011. So far, thirty-three isolates including 29 from Taiwan and 4 from Korea, were classified by eBURST as CC20 ( Table 5), suggesting that CC20 is an Asian cluster, which constitutes MLST clade 3. Interestingly, when comparing with CC4, another clade 3 cluster, CC20 isolates showed a significant expansion in north Taiwan.