Whole-Genome Sequence Analysis of Candida glabrata Isolates from a Patient with Persistent Fungemia and Determination of the Molecular Mechanisms of Multidrug Resistance

Whole-genome sequencing (WGS) was used to determine the molecular mechanisms of multidrug resistance for 10 serial Candida glabrata bloodstream isolates obtained from a neutropenic patient during 82 days of amphotericin B (AMB) or echinocandin therapy. For WGS, a library was prepared and sequenced using a Nextera DNA Flex Kit (Illumina) and the MiseqDx (Illumina) instrument. All isolates harbored the same Msh2p substitution, V239L, associated with multilocus sequence type 7 and a Pdr1p substitution, L825P, that caused azole resistance. Of six isolates with increased AMB MICs (≥2 mg/L), three harboring the Erg6p A158fs mutation had AMB MICs ≥ 8 mg/L, and three harboring the Erg6p R314K, Erg3p G236D, or Erg3p F226fs mutation had AMB MICs of 2–3 mg/L. Four isolates harboring the Erg6p A158fs or R314K mutation had fluconazole MICs of 4–8 mg/L while the remaining six had fluconazole MICs ≥ 256 mg/L. Two isolates with micafungin MICs > 8 mg/L harbored Fks2p (I661_L662insF) and Fks1p (C499fs) mutations, while six isolates with micafungin MICs of 0.25–2 mg/L harbored an Fks2p K1357E substitution. Using WGS, we detected novel mechanisms of AMB and echinocandin resistance; we explored mechanisms that may explain the complex relationship between AMB and azole resistance.


Introduction
Candida bloodstream infections (BSIs) are the most common nosocomial fungal infections and are associated with high rates of mortality [1,2]. Candida albicans is the most common species causing candidemia; however, the proportion of candidemia cases caused by non-albicans Candida (NAC) species (e.g., Candida glabrata, Candida parapsilosis, and Candida tropicalis) is increasing worldwide [3]. The increasing frequency of BSI isolates of NAC species is associated with many different factors such as antifungal drug exposure, catheter use, intensive care unit admission, age, and geographic distribution [2,4]. Recent increases in antifungal use have led to increasing azole resistance among BSI isolates of NAC species; the emergence of multidrug-resistant (MDR) Candida strains, such as Candida auris and and the evolutionary process based on the molecular mechanisms present in this clonal population of C. glabrata.

Fungal Isolates and Antifungal Susceptibility Testing
All 10 serial BSI isolates of C. glabrata from our previous report were assessed [9]. The patient was treated with micafungin for 23 days from hospital day (HD) 82 to 104, with AMB for 16 days from HD 105 to 120 and for 40 days from HD 124 to 163, and with caspofungin for 6 days from HD 121 to 126 and for 16 days from HD 147 to 162. The 10 C. glabrata isolates tested in this study were recovered serially from blood cultures between HDs 99 and 160 [9]. All isolates were stored at −70 • C in trypticase soy broth supplemented with 15% glycerol. The antifungal MICs of fluconazole, voriconazole, posaconazole, itraconazole, anidulafungin, caspofungin, and micafungin were re-determined using the Sensititre YeastOne ® system (Thermo Fisher Scientific, Waltham, MA, USA) whereas the antifungal MICs of AMB were determined using ETEST ® (bioMérieux, Marcy-l'Étoile, France). Two reference strains, Candida parapsilosis ATCC 22019 and Candida krusei ATCC 6258, were included in each antifungal susceptibility test as quality control isolates. The MIC interpretive criteria included species-specific Clinical and Laboratory Standards Institute  [12]. The epidemiological cutoff values (ECVs) proposed in CLSI M59-ED3 or European Committee on Antimicrobial Susceptibility Testing (EUCAST; AMB only) were used as MIC interpretive criteria for AMB (susceptible, ≤1 mg/L; resistant, >1 mg/L), voriconazole (susceptible, ≤0.25 mg/L; resistant, >0.25 mg/L), posaconazole (susceptible, ≤1 mg/L; resistant, >1 mg/L), and itraconazole (susceptible, ≤4 mg/L; resistant, >4 mg/L) [24,25]. In this study, MDR was defined as resistance to two or more classes (triazoles/echinocandins/polyenes) of antifungal drugs [18]. Therapeutic failure was defined as either persistence of Candida in the bloodstream despite >72 h of antifungal therapy or the development of breakthrough fungemia during treatment with the indicated antifungal agents for >72 h [8,26].

Whole-Genome Sequencing
DNA was extracted from 10 C. glabrata serial isolates (isolates 1-10) as described previously [27]. A library was prepared using a Nextera DNA Flex Kit (Illumina, San Diego, CA, USA) and sequenced as 150-bp paired-ends using the MiseqDx (Illumina) instrument. The sequencing matrix was extracted using Sequencing Analysis Viewer version 2.4.7 (Illumina). Adapter sequences and low-quality bases were trimmed using BBDuk from BBMap package version 38.95 [28]. The trimmed sequences were aligned to the reference genome using Burrows-Wheeler Aligner version 0.7.17 with the BWA-MEM algorithm [29]. The C. glabrata reference genome (CBS 138) from the Candida Genome Database was used as the reference genome [30]. Duplicate marking and conversion to the BAM file were performed using Picard version 2.27.4 [31]. Using HaplotypeCaller in GATK version 4.2.6.1, variants including single-nucleotide polymorphisms (SNPs) and insertions and deletions (INDELs) were called. Variants were filtered as described previously [32] and those that had a depth below 10 were removed. Annotation was carried out using snpEff version 4.3t with GFF file version 3 from the Candida Genome Database [30]. To filter the variants further, a list of genes associated with antifungal resistance was curated using the file CBS138 of chromosomal features from the Candida Genome Database [30] with the keywords '(drug) resistance', 'resistant', 'antifungal (target)', and '(suppress/reduced) sensitivity'. ERG genes, FCY1, FCY2, FKS3, HSP90 (HSC82), MMR1 (CAGL0A04169g), NDT80, TAC1 (HAL9), and UPC2 (UPC2A and UPC2B), which were missing from the initial keyword search but needed for further investigation of three classes of antifungal agents and MDR, were added to the list [33][34][35]. Thus, as the criteria for shortlisting MDR genes, we used 182 genes associated with antifungal resistance (Supplementary Table S1). These included the representative ERG genes, CDR1, CDR2 (PDH1), FEN1, FKS1, FKS2, FKS3, FLR1, SNQ2, PDR1, and QDR2 [36]. Additionally, phylogenetic analysis was performed based on SNP data for 182 resistance genes, using the maximum-likelihood method with the Kimura two-parameter model and bootstrap analysis with 1000 replications in MEGA version 11.0.11 [37]. Nonsynonymous mutations associated with antifungal resistance and correlated with the antifungal MICs were visually inspected using Integrative Genomics Viewer version 2.14.0 (Supplementary Figure S1). This study was approved by the Ethics Committee of Chonnam National University Hospital (CNUH) Gwangju, Korea; the need for informed parental consent was waived due to the retrospective nature of the study (CNUH-2014-290).

In Vivo Virulence Analysis Using Galleria Mellonella
We evaluated the virulence of five serial isolates (isolates 1-5) of C. glabrata and 35 BSI isolates of C. glabrata obtained from Korean multicenter surveillance cultures (18 FR isolates harboring Pdr1p mutations, and 17 F-SDD isolates without Pdr1p mutations) [8] in the G. mellonella insect model, as described previously [38,39]. Briefly, groups of 20 larvae (~150 mg; S-worm, Cheonan, Republic of Korea) were stored in wood shavings in the dark at 18 • C prior to use. The following three control groups were included: larvae injected with 10 µL of phosphate-buffered saline (N = 20), larvae that received needle injury only (N = 20), and untouched larvae (N = 20). A Hamilton syringe (25 gauge, 50 µL) was used to inoculate larvae with C. glabrata; it was also used to apply treatment or control solutions to the larvae. To determine the virulence of clinical C. glabrata isolates, larvae were infected with 5 × 10 6 conidia per larvae; survival was monitored up to 96 h post-infection at 37 • C. Data were combined to calculate the mean percentage survival.

Statistical Analysis
RStudio version 2022.7.1.554 (RStudio, Inc., Boston, MA, USA) was used for statistical analysis. The Wilcoxon rank-sum test or Student's t-test was used to determine the significance of between-group differences in survival at 24, 48, 72, and 96 h, based on the Shapiro-Wilk normality test and F-test. Differences were considered statistically significant at p < 0.05.

Deposition of the Raw Sequence Data
The raw sequence data were deposited in the NCBI Sequence Read Archive (BioProject PRJNA949257).

Results
In the WGS analysis, each run matrix was within the manufacturer's recommended value (Supplementary Table S2) [40]. An average of 5,222,511 reads were produced per isolate and 98.6% of the total reads were mapped to the reference genome (CBS138) with 55.4× to 70.8× coverage (average 61.6×). After variant calling, a total of 90,650 mutations (9601 INDELs and 81,049 SNPs) were detected per isolate; 12.8% were nonsynonymous mutations (Table 1). When filtered according to the resistance genes in which nonsynonymous mutations were detected, an average of 251 mutations (16 INDELs and 235 SNPs) were observed per isolate, of which 238 (94.8%) were simultaneously observed in all isolates. We detected 0.421 to 0.438 SNPs/kb among 10 isolates. Phylogenetic analysis based on WGS SNP data for the 182 resistance genes showed considerable diversity among 10 isolates, regardless of isolation date or antifungal susceptibility pattern (Supplementary Figure S2).
showing increased MICs against AMB (2-3 mg/L) harbored ERG6 R314K (isolate 5), ERG3 G236D and ERG4 P227fs (isolate 8), and ERG3 F226fs (isolate 10) mutations. With regard to azole resistance, a PDR1 GOF mutation (L825P) was observed in the 10 serial isolates. Six FR isolates (isolate 1, 2, 4, and 8-10) showed markedly high MICs (≥256 mg/L) for fluconazole and higher MICs for voriconazole, posaconazole, and itraconazole. However, the fluconazole MICs were 4-8 mg/L in four (isolate 3 and 5-7) isolates that harbored ERG6 A158fs or R314K simultaneously. These four isolates also showed lower MICs for other azoles. With regard to echinocandins, all of the isolates exhibited  Figure 2 depicts the possible evolution of the antifungal mechanisms of the 10 sequential clonal C. glabrata isolates with the same Pdr1p L825P mutation during the course of AMB or echinocandin therapy. All 10 isolates were associated with breakthrough fungemia during the administration of echinocandins (isolates 1 and 4), AMB (isolates 2, 3, and 5-8), or both (isolates 9 and 10). The Fks2p K1357E mutation first appeared after 17 days of micafungin exposure in isolate 1 and it was shared by five subsequent isolates (isolates 3, 6, 7, 8, and 10). These six isolates were designated as subpopulation #1. The Erg6p A158fs mutation first appeared after 16 days of AMB therapy in isolate 3 (subpopulation #1-2) and was shared by two subsequent isolates (isolate 6 and 7; #1-2). Clonal subpopulation #1 re-appeared after 6 days of caspofungin therapy and 35 days of AMB therapy with the addition of Erg4p P227fs combined with an Erg3p G236D mutation (isolate 8; subpopulation #1-3), and after 19 days of caspofungin and 52 days of AMB therapy with an additional Erg3p F226fs mutation (isolate 10; subpopulation #1-4). Overall, echinocandin breakthrough fungemia was caused by two isolates (isolate 1 and 10) of subpopulation #1, which harbored the Fks2p K1357E mutation. On the other hand, Fks2p I661_L662insF and Fks1p C499fs appeared in isolate 2 after 23 days of micafungin therapy; these mutations were shared by isolate 5, so isolates 2 and 5 were designated as clonal subpopulation #2. Clonal subpopulation #2 with an additional Erg6p R314K mutation appeared after 22 days of AMB therapy (isolate 5; subpopulation #2-2). In addition, two isolates (isolate 4 and 9) had a unique Fks2p mutation (designated as clonal subpopulations #3 and #4, respectively).  Table 2 shows the virulence in the G. mellonella model, among C. glabrata isolates 1 to 5. In vivo assays in the insect G. mellonella revealed that the 96-h survival rates of G. mellonella larvae infected with four isolates (isolates 1, 2, 3, and 5) were relatively higher than survival rates of G. mellonella larvae infected with FR or F-SDD isolates. The mean survival rate in larvae infected with FR isolates (N = 18) was significantly higher than the  Table 2 shows the virulence in the G. mellonella model, among C. glabrata isolates 1 to 5. In vivo assays in the insect G. mellonella revealed that the 96-h survival rates of G. mellonella larvae infected with four isolates (isolates 1, 2, 3, and 5) were relatively higher than survival rates of G. mellonella larvae infected with FR or F-SDD isolates. The mean survival rate in larvae infected with FR isolates (N = 18) was significantly higher than the rate in larvae infected with F-SDD isolates (N = 17) at all four time periods examined (24 h, P = 0.010; 48 h, P = 0.003; 72 h, P = 0.002; 96 h, P = 0.006).

Discussion
The development of resistance to C. glabrata BSI isolates during treatment is a possible cause of treatment failure, but few reports have provided a comprehensive understanding of how C. glabrata genomes can accumulate gene mutations that result in phenotypic resistance to antifungals during an extended course of antifungal therapy. In the present study, all 10 isolates harbored the same Msh2p substitution, V239L, which is known to be associated with both MLST type ST7 [8] and hypermutability [41]. All isolates harbored the same Pdr1p L825P mutation, which is associated with azole resistance [8]. Our WGS showed that the 10 isolates had a relatively low density of SNPs (0.421-0.438 SNPs/kb), reflecting their clonal nature, and that the genetic changes in antifungal drug-associated genes were due to long-term antifungal therapy [36,42]. An important implication of our findings is the high concordance between several nonsynonymous mutations in genes affecting AMB or echinocandin resistance and their MICs. For the first time, we have demonstrated that the presence of Erg6p mutations in C. glabrata isolates with Pdr1p GOF mutations could lower fluconazole MICs.
Acquired AMB resistance in Candida isolates is rare [43][44][45][46]. The rare occurrence of AMB resistance in C. glabrata may be partly due to a lack of detection ability using current CLSI or the European Committee on Antimicrobial Susceptibility Testing reference methods. In this study, the AMB MICs of six isolates were ≥2 mg/L by the ETEST ® , but those of all 10 isolates were 0.5-1 and 0.5-2 mg/L by the CLSI M27 method and Sensititre Yeast One ® system, respectively (data not shown), in agreement with a previous report [47]. The limited studies available suggested a mechanistic role for ERG2, ERG3, ERG4, and ERG6 in AMB resistance [44][45][46]48,49]. Previous studies reported a nonsense mutation [44] and missense mutation [50] in ERG6 that resulted in AMB resistance due to a composition change in sterol, which is the target of polyene. Here, we showed that three isolates of C. glabrata harboring a disruptive frameshift mutation (A158fs) in ERG6 exhibited markedly increased MICs (8-12 mg/L) for AMB and that harboring a substitution mutation (R314K) in ERG6 moderately increased the AMB MIC to 2 mg/L. Thus, ERG6 may be involved in AMB resistance in C. glabrata. An I207V mutation in ERG2 was also detected but was found in all isolates (isolates 1-10). Mutations in ERG3 or ERG4 have been found in AMB-resistant Candida albicans [48,51,52] and Saccharomyces cerevisiae [53], but rarely in C. glabrata [54]. In the present study, two isolates with AMB MICs of 2-3 mg/L harbored ERG3 G236D and ERG4 P227fs (isolate 8) and ERG3 F226fs (isolate 10) mutations, which may require more supporting evidence.
In our previous study, by comparing the PDR1 sequences of each C. glabrata isolate with the reference PDR1 sequence of C. glabrata (GenBank accession no. FJ550269) [55], we demonstrated that nearly all FR BSI isolates of C. glabrata in Korea harbored FR-specific Pdr1p mutations by excluding MLST genotype-specific Pdr1p amino acid substitutions [8].
In this study, all 10 isolates had an FR-specific nonsynonymous mutation (L825P) in PDR1, which may mediate azole resistance in C. glabrata [8]. However, among the isolates, four (isolate 3 and 5-7) showed low azole MICs (F-SDD) despite a PDR1 GOF mutation (L825P), while six had a fluconazole MIC ≥ 256 mg/L (FR). All four F-SDD isolates harbored an Erg6p (A158fs or R314K) mutation. A previous study showed that the lower ergosterol content associated with a nonsense mutation in ERG6 may have an indirect effect on susceptibility to azoles by preventing the targeting of efflux pumps to the plasma membrane, thereby favoring the accumulation of these drugs within the cell [44]. The presence of ERG6 mutations could lead to defects in ergosterol synthesis and changes in the binding of the efflux pump. In our previous study, FR isolates of C. glabrata exhibited higher mean expression levels of CgCDR1, CgCDR2, and CgSNQ2, compared with F-SDD isolates [8].
When we compared the expression levels of CgCDR1, CgCDR2, and CgSNQ2 in five isolates with the same Pdr1p L825P mutation, without (isolates 1, 2, and 4; FR isolates) or with (isolates 3 and 5; F-SDD isolates) an Erg6p mutation, the expression levels of CgCDR1 and CgSNQ2 in the three FR isolates were relatively higher than levels in the two F-SDD isolates, and similar to the mean expression levels of CgCDR1 and CgSNQ2 in 30 FR isolates harboring Pdr1p mutations. Taken together, our findings indicate that C. glabrata isolates with the same Pdr1p GOF mutations do not always show the same FR result-they can be F-SDD in AMB-resistant isolates with Erg6p mutations.
Although sequencing of the HS regions in FKS genes is the most convenient way of determining echinocandin resistance mechanisms, mutations occurring outside of these HS regions can also lead to echinocandin therapeutic failure, which confirms the importance of sequencing the entire FKS gene [4,56]. In the present WGS analysis, four isolates (isolate 2, 4, 5, and 9) showed disruptive INDELs in HS regions of FKS2, and all isolates (except isolate 9) showed missense or frameshift mutations occurring outside of these HS regions. Of the isolates, two with anidulafungin, caspofungin, and micafungin MICs > 8 mg/L harbored not only the mutation Fks2p I661_L662insF but also Fks1p C499fs. Given that mutations in FKS1 or FKS2 [26,57], or the combination of a null function mutation in FKS1 and point HS mutation in FKS2 [56], could lead to strong resistance among C. glabrata strains, the unique HS mutations (I661_L662insF and F659del) in FKS2 found in this study may have different impacts on echinocandin MICs according to the combination of other FKS nonsynonymous mutations. There were also two isolates harboring the mutation FKS2 F659del with or without the upstream FKS2 S201fs mutation. Relatively strong resistance to echinocandins was observed in an isolate harboring a single F659del mutation in FKS2 (isolate 9), but the echinocandin MIC was slightly decreased in an isolate harboring both mutations (F659del and S201fs) in FKS2 (isolate 4). The reasons for the lowered echinocandin MIC in isolate 4 harboring the F659del and S201fs mutations are uncertain. One possibility is that the upstream FKS2 S201fs mutation may affect the downstream FKS2 F659del mutation, but more evidence is needed. Six isolates in subpopulation #1 harboring the Fks2p K1357E mutation (isolates 1, 3, 6-8, and 10) showed micafungin MICs of 0.25-2 mg/L. The Fks2p K1357E mutation first appeared as breakthrough fungemia after 17 days of micafungin exposure in isolate 1, and five additional isolates were recovered from HD 99 to HD 160 despite further micafungin (5 days) or caspofungin (19 days) therapy. The role of Fks2p K1357E in echinocandin resistance remains uncertain as this SNP has not been described previously. However, a previous report showed that C. glabrata isolates harboring a single non-HS mutation in an FKS gene showed slightly increased MICs for echinocandins [58]. Here, we found that the echinocandin breakthrough fungemia was caused by two isolates (isolate 1 and 10), suggesting that the Fks2p K1357E mutation is associated with echinocandin therapeutic failure. Overall, our WGS study suggests that isolates harboring nonsynonymous mutations located outside the HS regions in FKS genes can increase echinocandin resistance.
C. glabrata BSI isolates from a particular geographic area have been reported to comprise a small number of major STs, according to MLST analysis. MLST of Korean BSI isolates showed that ST7 (47.8%) was the most common type, followed by ST3 (22.5%); the remaining isolates exhibited 28 types of minor STs [59]. FR isolates of C. glabrata typically had one Pdr1p amino acid substitution, which were rarely shared by two isolates from the same hospital in the same year, in agreement with a previous report that C. glabrata transmission between patients is rare [60]. Although our isolates exhibited ST7, the most common ST in Republic of Korea, none collected in 2009-2018 harbored Pdr1p L825P mutation except our 10 isolates, suggesting independent development of FR in C. glabrata in most patients [8,60]. Despite the clonal nature of the BSI isolates of C. glabrata obtained from our patient, the serial isolates showed significant non-serial phenotypic MIC variations to AMB, azole, or echinocandins. Similarly, phylogenetic analysis by WGS showed substantial genetic diversity, regardless of isolation date and phenotypic antifungal susceptibility pattern (Supplementary Figure S2).
We postulated that subpopulations with different resistance profiles are likely to have persisted in the gut and alternately invaded the bloodstream under selective pressure, highlighting the adaptability of C. glabrata to long-term treatment with various antifungal agents [9]. In the present study, WGS enabled us to detect the possible molecular mechanism responsible for the low-and high-level antifungal resistance of each isolate of C. glabrata and to show the evolution of molecular mechanisms within the same subpopulation due to different resistance profiles. Our findings suggest that some nonsynonymous mutations found in the same subpopulation (subpopulations #1 and #2) may represent pre-existing mutations, and some new mutations occurred after antifungal drug exposure. Subpopulations with pre-existing mutations are likely to persist in the gut or other mucosal sites and appear in the bloodstream with or without new genetic changes during the long course of antifungal therapy.
The fitness cost related to antifungal resistance acquisition by C. glabrata is unclear, and few studies have been reported thus far [61][62][63]. G. mellonella has been used as a host model to study C. glabrata virulence and antifungal efficacy [64]. In the present study, the mean survival rate in larvae infected with FR isolates was significantly higher than the rate in larvae infected with F-SDD isolates at all four time periods examined, indicating that F-SDD isolates without PDR1 mutations may be more virulent than FR isolates harboring PDR1 mutations. Moreover, our results suggest that our MDR C. glabrata isolates with PDR1 gene mutations have reduced virulence in the G. mellonella model.
A notable limitation of this study is that, although many of the detected mutations were located in genes involved in resistance, we did not directly assess their roles in resistance. The reintroduction of mutant alleles into susceptible strains via gene editing would be a useful approach for determining their roles in resistance. We used WGS to detect specific genetic alterations associated with antifungal resistance in serial clonal C. glabrata isolates. Some of these newly detected mutations were out of the target region in the gene (e.g., non-HS regions in FKS1/2) or out of the target gene (e.g., ERG genes) from our previous study based on conventional sequencing [9]. In this study, novel HS mutations (F659del mutations) were detected by WGS in isolates 4 and 9.

Conclusions
In conclusion, this study provides important perspectives on the utility of WGS for detecting molecular mechanisms of multidrug resistance based on 10 serial C. glabrata BSI isolates obtained from a patient with breakthrough fungemia during extended AMB or echinocandin therapy. Pdr1p GOF and Fksp mutations in C. glabrata may not always have the same effects; they may cause different levels of antifungal resistance, depending on the combination of nonsynonymous mutations present. Fluconazole MICs are lower in C. glabrata isolates with the same Pdr1p GOF mutation than in AMB-resistant isolates with Egr6p mutations. Fks2p HS mutations combined with Fks1p null-function mutations contribute to high-level echinocandin resistance. In addition, Fks2p mutations outside HS regions contribute to low-level echinocandin resistance. Persistent subpopulations of C. glabrata undergoing continuous clonal genetic evolution during long-term antifungal therapy could be responsible for the non-serial multiple antifungal resistance phenotypes of C. glabrata BSI isolates. In conclusion, WGS will improve the detection and monitoring of molecular mechanisms of antifungal resistance.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/jof9050515/s1, Figure S1: An example of Integrative Genomics Viewer showing mutations on FKS2 for 10 serial isolates of C. glabrata; Figure S2: Phylogenetic analysis of 10 serial C. glabrata isolates using SNP data for 182 resistance genes. Scale bar, number of nucleotide substitutions per site; numbers on nodes, bootstrap resampling values; Table S1: The list of possible genes associated with antifungal resistance in C. glabrata; Table S2: Sequencing and post-sequencing parameters of whole-genome sequencing in this study.  Informed Consent Statement: Patient consent was waived due to the retrospective nature of the study.

Data Availability Statement:
The raw sequencing data were deposited in the NCBI Sequence Read Archive (bioproject PRJNA949257).

Conflicts of Interest:
The authors declare no conflict of interest.